Chapter 3 Statistical Estimation of The Regression Function

Size: px
Start display at page:

Download "Chapter 3 Statistical Estimation of The Regression Function"

Transcription

1 Chapter 3 Statistical Estimation of The Regression Function 3.1 Statistical Estimation If the population can be observed, there is no statistical problem: all the features if the population are known. The problem of statistical inference arises when the available information consists of a limited sample that is randomly drawn from the (possibly infinitely large) population and we want to infer something about the population using the sample at hand. Statistical estimation is one aspect of statistical inference 1 - it concerns the estimation of population parameters such as the population mean and variance and the coefficients of a linear regression function. In Chapter 2 the population regression function of Y given X is defined as the conditional mean function of Y given X, written as E(Y X). An important reason for our interest in this functional relationship is that it allows us predict Y given values of X and to quantify the effect of a change in X on Y (measured by say the derivative with respect to X.) Moreover, the conditional predictions of Y that are produced by the regression function are optimal under specific but generally applicable conditions. This chapter is concerned with the problem of estimating the population regression function using a sample drawn from the population Parametric Versus Nonparametric Methods Figure 2.7 and the related discussion illustrates how a sample can be used to estimate a population regression. Since the population regression function of Y given X is the conditional mean of Y given X, we simply computed a sequence of conditional means using the sample and plotted them. Nothing in the procedure constrains the shape of the estimated regression. Indeed, the empirical regression of Size given Price (the plot of in Figure 2.7) wanders about quite irregularly (although as it does so it retains a key feature that we expect of the population regression of S given P, namely that its average slope is steeper than the major axis - the empirical regression starts off below the major axis and then climbs above it.) The method used to estimate the empirical regression functions in Figure 2.7 variables. 1 Two other important inference problems are: hypothesis testing and the prediction of random

2 Econometrics Text by D M Prescott Chapter 3, 2 can be described as nonparametric. While there is a huge literature on nonparametric estimation, this book is concerned almost entirely with parametric models. To illustrate the distinction between parametric and nonparametric methods, consider the equation Y = a + bx. This equation has two parameters (or coefficients): a and b and clearly the relationship between Y and X is linear. By varying the values of a and b the line s height and slope can be changed, but the fundamental relationship is constrained to be linear. If a quadratic term (and one more parameter) is added: Y = a + bx + cx 2, the relationship between Y and X becomes more flexible than the linear function. Indeed, the quadratic form embraces the linear form as a special case (set c = 0). But the linear form does not embrace the quadratic form: no values of a and b can make the linear equation quadratic. Of course, the three parameter quadratic equation is also constrained. A quadratic function can have a single maximum or a single minimum but not both. Quadratic functions are also symmetric about some axis 2. If further powers of X are added, each with its own parameter, the relationship becomes increasingly flexible in terms of the shape it can take. But as long as the number of parameters remains finite, the shape remains constrained to some degree. The nonparametric case is paradoxically not the one with zero parameters but the limiting case as the number of parameters increases without bound. As the number of terms in the polynomial tends to infinity, the functional relationship becomes unconstrained - it can take any shape. As noted above, the method used to construct the empirical regressions in Figure 2.7 did not constrain the shape to be linear, quadratic or any other specific functional relationship. In that sense the method used in Chapter 2 to estimate the population regression can be called nonparametric. In the context of regression estimation, the great appeal of nonparametric methods is that they do not impose a predetermined shape on the regression function - which seems like a good idea in the absence of any information as to the shape of the population regression. However, there is a cost associated with this flexibility and that concerns the sample size. To perform well 3, the nonparametric estimator generally requires a large sample (the empirical regressions in Figure 2.7 used a sample of almost 5,000 observations). In contrast, parametric methods that estimate a limited number of parameters can be applied when samples are relatively small. The following examples by-pass the statistical aspect 2 The graph of Y = a + bx + cx 2 is symmetric about the line X = - b/(2c) 3 The meaning of performing well will be discussed later in the chapter.

3 Econometrics Text by D M Prescott Chapter 3, 3 of the argument but nevertheless provide some intuition. If you know that Y is a linear function of X, then two points (2 observations) are sufficient to locate the line (and to determine the two parameters.) If you know the relationship is quadratic, just three points are sufficient to plot the unique quadratic function that connects the three points and therefore three observations will identify the three parameters of the quadratic equation. The relationship continues: in general n points will determine the n parameters of an n th order polynomial. 3.2 Principles of Estimation As discussed in Chapter 2, there are examples of bivariate distributions in which the population regression functions are known to be linear. In the remainder of this chapter we will be concerned with linear population regressions and the methods that can be used to estimate them. We begin with a discussion of alternative approaches to statistical estimation - all of which are parametric The Method of Moments The quantities are referred to as the first, second and third uncentred moments of the random variable X. The centred moments are measured around the mean The Method of Moments approach to estimating these quantities is to simply calculate their sample equivalents, all of which take the form of averages. Table 3.1 provides the details for the first two moments. Notice the parallels between the expressions for the population moments and their sample counterparts. First, the estimator uses instead of the expectation operator E. Both take an average, one in the sample, the other in the population. Second, the estimator is a function of the observations X i whereas the population moment is defined in terms of the random variable X.

4 Econometrics Text by D M Prescott Chapter 3, 4 Population Moment (parameter) Method of Moments Estimator Table 3.1 The justification for the Method of Moments approach to estimation is based on a Law of Large Numbers 4 which, loosely, states that as the sample size tends to infinity the probability that the sample mean differs from the population mean tends to zero. In other words, the probability limit of the sample mean is the population mean. In fact, the probability limit of any sample average is the expected value of that quantity. In the following expressions plim refers to the probability limit. Recall that the sample covariance is also a sample average, so this too is a consistent estimator of the population covariance. An estimator whose probability limit is identical to the parameter it estimates is said to be consistent. By the Law of Large Numbers, the Method of Moments (MM) estimator is a consistent estimator. An important property of the probability limit is provided by the following theorem: Theorem 3.1 If is a consistent estimator for the population parameter and f( ) is a continuous function then 4 See the Appendix to this chapter for more details on the Law of Large Numbers and the notion of probability limit

5 Econometrics Text by D M Prescott Chapter 3, 5 Theorem 3.1 implies for example that whereas. Now let s apply the MM estimator to the bivariate linear regression. Table 3.2 presents the details; they are based on Theorem 2.1 of Chapter 2. That theorem states that for any linear population regression E( Y X ), the slope and intercept are given by. The MM estimator is simply the sample counterpart to the expression that defines the population parameter of interest. Table 3.2 The Method of Moments Estimator for The Bivariate Linear Regression: E(Y X) Population Parameters MM Estimator Slope:. Intercept: Later in this chapter we will report the MM estimator for the linear regression of Price on Size using the house-price data that were discussed in Chapter The Maximum Likelihood Estimator An important difference between the Maximum Likelihood Estimator (MLE) and the MM estimator discussed in the previous section is that the MLE demands that the specific distribution that generated the data be identified. In the previous section we assumed that the population regression is linear, but we did not specify or assume that the random variables X and Y are, for example, bivariate normal. If it is known 5 that X and Y are bivariate normal, then intuitively, it seems sensible to take this into account when estimating the parameters of the regression function. An important property shared by 5 In practical situations it can rarely be known with certainty what distribution actually generated the data but through various tests the statistician may be comfortable assuming that X and Y are, say, normally distributed

6 Econometrics Text by D M Prescott Chapter 3, 6 MM and ML is that both estimators are consistent. In the context of MLE, the researcher is assumed to know the distribution from which the data are drawn - as noted in Chapter 1, it may be helpful to think of this distribution as a data generating process in the way that rolling a six-sided die generates data. The principle behind MLE is essentially this: given the data, what are the values of the population parameters that make the observed sample the most likely. That is, what kind of population is likely to have generated this particular sample? Suppose there are two colleges - one specializes in sports the other in music. Suppose the population mean height of female students at these colleges is 1.70 metres (sports) and 1.63 metres (music). A random sample of 20 students is taken from one of the colleges and the sample mean height is 1.64 metres. From which college was the sample drawn? It might never be known for certain, but the music college is more likely than the sports college 6 to generate a sample mean of The ML principle identifies the music college as the source of the sample. Now consider a more formal example that illustrates how the ML principle is applied. Suppose the object is to estimate the proportion of grade 10 students that smoke cigarettes. In the population of grade 10 students the true proportion is B. A random sample of size n reveals that n 1 smoke and n 0 do not. The probability of observing n 1 smokers and n 0 non-smokers in a sample of n = n 1 + n 0 is given by the binomial distribution: where k is the binomial coefficient The MLE treats the sample as given (n and n 1 are thought of as fixed) and asks what value of B makes the actual sample most likely (most probable in this case.) Let be the MLE and is some other value. The MLE satisfies: The value of can be found using calculus: take the derivative with respect to B of the 6 We are implicitly assuming the variance of height is the same at the two colleges.

7 Econometrics Text by D M Prescott Chapter 3, 7 probability of observing the sample and set it to zero 7. The solution is = n 1 /n, namely the proportion of smokers in the sample. The MLE of B is therefore perfectly intuitive: the proportion of smokers in the population is estimated by the proportion of smokers in the sample. To apply the ML principle to the bivariate regression model it is necessary to specify the distribution that generated the data, such as the bivariate normal. Equation [2.9] of Chapter 2 describes the regression of Y given X for the bivariate normal distribution. It is reproduced here: where, is a normally distributed random variable with a mean of zero and a variance. The normal density function is given be equation [1.13]. For the random variable, it has the form The sample consists of n observations (X i, Y i ), i = 1, 2,..., n. The corresponding values of are not observable, but nevertheless the likelihood of observing the sample can be expressed as the product of the densities: Recall that the exp(a)exp(b) = exp(a+b) - the product of exponentials is the exponential of the summed exponents. Apply this idea to the likelihood function and we get The final step is to substitute for the unobserved, s using [2.9]. This expresses the likelihood of the sample in terms of observable data: 7 Treat the probability as a product i.e., use the product rule of differentiation. It can be shown that the first order condition identifies a maximum - not a minimum.

8 Econometrics Text by D M Prescott Chapter 3, 8 In equation [3.3], X and Y represent the n observed values (X i, Y i ), i = 1, 2,..., n.. Note also that the likelihood function is seen as a function of the unknown parameters. The ML estimators are the parameter values that maximize the likelihood function, treating X and Y as fixed. If are any other parameter values then the MLE satisfies As in the previous example, calculus can be used to determine the MLE. The details are omitted and we go straight to the solution. It turns out that in this case the MLE of " and $ are identical to the MM estimators given in Table Ordinary Least Squares In Chapter 2 it was explained why the population regression function can be described as the least squares function. The argument is briefly reviewed here. If the object is to find the best predictor of the random variable Y, say Y *, such that Y * minimises, the solution is Y * = E(Y), the population mean of Y. Further, if X is correlated with Y and the value of X is known when the forecast of Y is made, then the solution to the optimal prediction problem is Y * = E(Y X) i.e., the conditional mean of Y given X. This is none other than the population regression function of Y given X. The regression function therefore minimises the expected (long run average) squared prediction error. Consider now what this implies if the population regression function is linear. In such a case it can be written as Let Y* =. + 0X be a representative linear equation. Consider the problem: Determine the values of. and 0 that: minimise E (Y - Y* X) 2 = E ( Y X X) 2

9 Econometrics Text by D M Prescott Chapter 3, 9 Since we know the solution to this minimisation problem is the population regression function E(Y X) and since in this case it is linear, the solution values are:. = " and 0 = $ where " and $ are the specific parameter values defined above. This analysis suggests that the least squares line drawn through a sample scatter is a viable estimator of the linear population regression function. Table 3.3 compares the properties of the linear population regression with the sample least squares regression. The sample of n observations is represented by the points (X i, Y i ) for i = 1,2,...,n. Table 3.3 The Least Squares Method of Estimation The Population The Sample Y = " + $X +, Y i = a + b X i + e i for i = 1,2,...,n. Linear Population Regression Estimated Linear Regression " and $ are the unique values that minimise: The Least Squares Principle: Choose a and b such that the following quantity is minimised: Population Parameter Values The Least Squares Solution Values Table 3.3 emphasises that the population regression function is the least squares function. To estimate the parameters of this function using a sample that has been drawn from the population we find the least squares function within the available sample. Notice that to apply the least squares principle the

10 Econometrics Text by D M Prescott Chapter 3, 10 expectation operator E (which gives a population mean) is replaced by its sample equivalent (which gives a sample mean). Before looking at the details of how the least squares solution is obtained, consider the numerical example in Table3.4 Table 3.4 Annual Advertising and Sales Data for Eight Stores (Thousands of Dollars) Store No Advertising Expenditures Sales The artificial data in Table 3.4 represent the sales for eight stores (the dependent or Y-axis variable) together with each store s advertising expenditure (the explanatory or X-axis variable.) The data are plotted in Figure 3.1 along with the least squares regression line. Figure 3.1 The Least Squares Regression Y^ = *18 = 980 Sales e = -50 (X=18,Y=930) Y^ = a + bx Advertising Expenditure The least squares equation is written as: $Y = a + bx. For each data point, the vertical distance from to the least squares line is referred to as the least squares residual, which is represented by the

11 Econometrics Text by D M Prescott Chapter 3, 11 symbol e. For the i th data point (X i, Y i ), the least squares residual is The least squares residual can also be described as a within-sample prediction error since it is the difference between the observed value of Y and the predicted value of Y, that is, the value predicted by the least squares regression equation. The equation of the least squares regression in Figure 3.1 is $Y = X. In thousands of dollars, store number 4 spent 18 on advertising and had sales of 930. The L.S. regression line predicts sales of *18) = 980 (thousands of $). The prediction error is therefore (thousands of $). Notice that all the data points below the L.S. regression line have negative residuals since they are over-predicted by the L.S. regression while all points above the line have positive residuals Solving the Least Squares Problem The slope and the intercept of the L.S. regression line are chosen in such a way as to minimise the sum of squared residuals, SSR. If the slope and intercept are changed, the residuals will obviously change as well and so too will the sum of squared residuals, SSR. In short, SSR is a function of a and b which can be written as follows: To solution to the L.S. minimisation problem can be found by setting to zero the first derivatives of SSR(a,b) with respect to a and b. The pair of first order conditions provide two equations that determine the solution values of a and b. The two partial derivatives are shown in equations [3.4] and [3.5]

12 Econometrics Text by D M Prescott Chapter 3, 12 Cons ider equation [3.4] first. Notice that the differential operator can pass through the summation sign because the derivative of a sum of items is the same as the sum of the derivatives of the individual items. Equation [3.4] can be written as: The derivative of the typical element can be evaluated in one of two ways. Either the quadratic term can be expanded and then differentiated, or the function of a function rule can be used. Using the function of a function rule we find that Since the derivative is set to zero at the minimum point of the function S(a, b), the term -2 can be

13 Econometrics Text by D M Prescott Chapter 3, 13 cancelled. The first order condition [3.4] can therefore be written as Equation [3.6] has an interesting interpretation that will be discussed later. The final step is to rewrite [3.6] in a more useable form. Recall that the sum of n numbers is always equal to n times the mean of the numbers: Also, recall that The final form of the first order condition [3.4] is Now consider the second of the first order conditions, [3.5]. The derivative of the typical element with respect to the slope coefficient b is Equation [3.5] can therefore be written as Dividing through by minus two yields the following equation that is equivalent to equation [3.6].

14 Econometrics Text by D M Prescott Chapter 3, 14 After some rearrangement, equation [3.8] implies that the least squares coefficients satisfy: Equations [3.7] and [3.9] can be solved for the least squares coefficients. Equation [3.7] is used to solve for the intercept a. Now substitute for a in equation [3.9] and solve for b: Notice that the least squares equation for the slope coefficient b can be expressed in deviation form, where A Numerical Example Table 3.5 illustrates the calculation of the least squares coefficients for the advertising/ sales data that are plotted in Figure 1. The first two columns present the original data on advertising expenditure and sales at the two stores. The least squares slope coefficient b is calculated according to equation [4.13], which requires the computation of GX i Y i and G(X i ) 2 as well as the means of X and Y. The squared X values appear in the third column and the cross products between X and Y appear in the fourth column of Table 3.5. These sums and the means of X and Y are presented at the bottom of the appropriate columns. Finally, the least squares formulae are used to compute the intercept and slope of the least squares line for these data. These calculations show that the line drawn in Figure 1 is indeed the least squares line.

15 Econometrics Text by D M Prescott Chapter 3, 15 Table 3.5 Calculation of the Least Squares Coefficients (Advertising) X i (Sales) (X i ) 2 X i Y i Y i (GX i )/n = X - = 20 (GY i )/n = Ȳ= 1000 G (X i ) 2 = 3626 G X i Y i = Interpretation of the L.S. Regression Coefficients The parameter $ in Table 3.3 can be described as the slope of the population regression function E(Y X) i.e., it is the derivative of E(Y X) with respect to X. A more intuitive $ interpretation is that it represents the effect on the conditional mean of Y, E(Y X), of a unit change in X. $ therefore has units which are equal to the units of Y divided by the units of X. The L.S. estimator b estimates $ and so it is

16 Econometrics Text by D M Prescott Chapter 3, 16 the estimated effect on E(Y X) of a unit change in X. Table 3.6 shows the L.S. regression of house price on a constant and house size. These data have been described in Chapter 2. Recall that the data were collected over a six year period, The variable price records the price at which the house sold and size is its size in square feet. The coefficient on SIZE is $60.5 per square foot and it represents the estimated effect on market price of an increase in SIZE of one square foot. More specifically, it is the effect on the conditional mean price (conditional on size) of a unit increase in size. Consider the population mean price of all houses that are exactly (a) 1500 square feet and (b) 1501 square feet. The difference in these conditional means is estimated to be $60.5 per square foot. Note that the relationship between the conditional mean price and size is linear so this estimate applies over the entire range of house sizes. However, it is best to think of the estimate as being particularly relevant to at the sample mean size, since this is where the weight of the data is concentrated (the balance point of the size distribution.) Also, since the data were collected over a period of 6 years when house prices were rising, it would be appropriate to think of the estimate of $ as applying at a date in the middle of the sample period, say January Dependent variable: PRICE Number of observations: 2515 Mean of dep. var. = Std. dev. of dep. var. = Table 3.6 The Least Squares Regression of Price on Size Estimated Variable Coefficient C SIZE The intercept of the L.S. regression is $15, Note that the intercept has the same units as the dependent or Y-axis variable which is PRICE in this case. In most L.S. regressions the intercept has no meaningful interpretation. On the other hand it is usually important to include the intercept in the equation otherwise the estimated linear relationship between Y and X will be forced through the origin

17 Econometrics Text by D M Prescott Chapter 3, 17 (0, 0) and this is rarely justified. It could be argued that in the current example, the predicted price of a house of zero size refers to the price of an empty lot. However, since the sample did not include any market transactions in which empty lots were bought and sold it is unlikely that the value of $15, is a particularly good estimate of the market value of an empty lot in say January L.S. chooses a slope and intercept to fit the data and the resulting linear equation is an approximation to the population regression over the range of the available data. In this case the scatter plot is a long way from SIZE = 0. What is meant by a long way? Table 3.7 shows that the minimum SIZE in the sample is 700 square feet and the standard deviation of SIZE is 392 square feet so SIZE = 0 is 1.8 standard deviations below the minimum size in the sample and 3.4 standard deviations below the sample mean of SIZE. It is extremely important to bear in mind that the interpretation of a particular regression coefficient depends crucially on the list of explanatory variable that is included in the regression. To illustrate this important point consider a model in which there are two continuous explanatory variables. To make the example specific, you might think of X 1 as house size and X 2 as lot size. The coefficient $ 1 is the partial derivative of E(Y X 1, X 2 ) with respect to X 1. It is therefore the effect of a change in X 1 on the conditional mean price while holding X 2 constant This holding X 2 constant is a new condition that did not apply when X 2 was not in the model. To make this point clear, compare the conditional mean of Y at two values, say X 1 and X The change in the conditional mean is The important point to note is that the terms cancel only if takes the same value in the two conditional means. When we consider the coefficient we are therefore comparing the mean price in two subpopulations of houses that have the same lot size but differ in house size by one square foot. We now turn to a model of house prices in which there are several explanatory variables. The definitions of these variables is given in Table 3.7

18 Econometrics Text by D M Prescott Chapter 3, 18 Table 3.7 Variable Definitions Symbol Description & Units PRICE Transaction price, ($) SIZE LSIZE AGE BATHP House size (square feet) Lot size (square feet) Age of house at time of sale (years) Number of bathroom pieces POOL If pool exists, POOL = 1, otherwise POOL = 0 SGAR If single-car garage, SGAR =1, otherwise SGAR =0 DGAR If double-car garage, DGAR =1, otherwise DGAR =0 FP If fireplace exists, FP = 1, otherwise FP =0 BUSY_RD BUSY_RD = 1 if on busy road, otherwise BUSY_RD = 0 T Time of sale to nearest month. T =1 if Jan. 83; T = 2 if Feb. 83 etc. Table 3.8 reports summary statistics for the variables described on Table 3.7 and the L.S. regression of PRICE on ten explanatory variables plus a constant term, which allows for the intercept. The coefficient on SIZE is $41.22 per square foot which is just 2/3 the value of the corresponding coefficient in Table 3.6 The reason for this substantial difference is that the regression in Table 3.6 conditions only on SIZE. But in Table 3.8 PRICE is conditioned on a much longer list of variables. From Table 3.8 we infer that a one square foot increase in size increases the mean price of houses by $41.22 while holding constant the lot size, the age of the house, the number of bathroom pieces and so on. If you were to walk round your neighbourhood, you will probably find that bigger houses are likely to be on bigger lots, have more bathroom pieces and perhaps have a double rather than a single garage. This is reflected in the larger L.S. regression coefficient on SIZE in Table 3.6 compared to that in Table 3.8. Now let s turn to a few other coefficients in Table 3.8. The coefficient on LSIZE is positive, which confirms larger lots add to the market value of houses. On the other hand, older houses sell for

19 Econometrics Text by D M Prescott Chapter 3, 19 less. The coefficient on AGE suggests that for every additional 10 years since construction, house prices fall by $1,891. Being on a busy road reduces the expected price by $3,215 while a fireplace is estimated to add $6,672 to market value. The coefficient on T is $1,397 which provides an estimate of how quickly prices were rising over the sample period, T records the month in which the transaction took place so an increase in T of 1 means one month has passed. The data suggest that house prices rose $1,397 per month over this six year period. Note that the model is linear in T so it estimates an average monthly increase - a linear time trend in prices. The model as it stands cannot reveal if prices rose slowly at first and then accelerated or rose quickly and then slowed. We would a need a more sophisticated model to track anything other than a linear price path. Table 3.8 Number of Observations: 2515 Mean Std Dev Min Max PRICE SIZE LSIZE AGE POOL BATHP FP SGAR DGAR BUSY_RD T Dependent variable: PRICE Number of observations: 2515 Mean of dep. var. = Std. dev. of dep. var. = Estimated Variable Coefficient C SIZE LSIZE AGE POOL BATHP FP SGAR DGAR BUSY_RD T Effects of Rescaling Data on The Least Squares Coefficients

20 Econometrics Text by D M Prescott Chapter 3, 20 In the previous section it was argued that a complete discussion of the least squares must include the units in which the variables are measured. This section presents two rules that show precisely how rescaling the data affects the least squares intercept and slope coefficient. The dependent and independent variables are quantities that have two parts: one component is the numerical part that is perhaps stored in a data file or on paper. The other component is the unit of measurement. Consider a small town with a population of 25,000 people. Clearly the "population" has two parts, a pure number (25,000) and the units (a person). In symbols: Quantity = (number) times (units). The same quantity can be expressed in different ways, for example we may prefer to reduce the number of digits we write down by recording the population in thousands of people: now Quantity = (25) times (thousands of people). Notice that this rescaling can be expressed as Quantity = (number/1000) times (1000xunits). The number component is divided by 1000 and the units component is multiplied by 1000 (the units are transformed from people to thousands of people). In the equation Y = a + bx, X and Y refer only to the number components of the relevant quantities, which is why it is so important to be aware of the units of measure. First consider rescaling the number component X by a scale factor m x ; define the result to be X * = m x X. Although X * and X are different numbers they represent the same quantity. Quantity = X times (units of X) = (m x X) times (units of X divided by m x ) = X * times (units of X * ). The units of X * are the units of X divided by m x. Replacing X with X * in the equation Y = a + bx will result in a new slope coefficient b *. Notice that Y is simply a number and it will not change as a result of this substitution so the new right hand side (a + b * X * ) must give the same result. The intercept a remains unchanged and the product of the slope and the X-axis variable, b * X *, is the same as before, i.e., b * X * = bx. The previous equation implies that the new slope coefficient is b * = b(x/x * ) = b/m x. The effect of rescaling the X-axis data is summarized in the following rule. Rescaling Rule #1 If the X-axis data are rescaled by a multiplicative factor m x, the least squares intercept is unchanged but the least squares slope is divided by m x. This rule illustrated by the following example. Suppose that the advertising data had been recorded in dollars instead of thousands of dollars but sales continue to be recorded in thousands of dollars. For example, store #1 spent $15,000 on advertising so instead of recording 15, suppose 15,000

21 Econometrics Text by D M Prescott Chapter 3, 21 appeared in Table 3.4. The slope of the least squares line can be recomputed using the method presented in Table 3.5 or a computer program such as TSP could be used. The result will be Y = X *, where X * is advertising measured in dollars. The new slope coefficient of 0.01 still represents the effect of a one unit increase in advertising expenditures on sales. A one dollar increase in advertising leads to a sales increase of (0.01)x(units of sales) = 0.01x$1 000 = $10. The basic conclusion remains in tact and is entirely independent of the units that the data are measured in. 8 The effects of rescaling Y by a multiplicative factor m y can be worked out in a similar way. When Y is multiplied by m y we obtain Y * = m y Y. Using Y * to compute the least squares line instead of Y we multiply the original least squares equation by m y : Y * = m y Y = (m y a) + (m y b)x. In this case, both the intercept and the slope coefficient are multiplied by m y. Rescaling Rule #2 If the Y-axis data are rescaled by a multiplicative factor m y, both the least squares intercept and slope coefficient are multiplied by m y. To illustrate this rule suppose that the sales data are measured in dollars while advertising figures continue to be measured in thousands of dollars. This change would cause all the numbers in the last row of Table 3.4 and all the numbers in the second column of Table 3.5 to be multiplied by m y = If you work through the calculations in Table 3.5 using the new numbers you will find that the new intercept coefficient is , i.e., the previous intercept is multiplied by Also, the new slope coefficient is it too is increased by a factor of The new least squares equation is Y^ = X. Again, the rescaling does not make any substantive change to the interpretation of the fitted line. A one unit increase in advertising expenditures ($1 000) raises sales by times (units of Y), which amounts to a $ increase in sales since the units of Y are simply dollars. Also, the predicted sales for store #1 are $ (10 000)($15) = $ , just as before Some Important Properties of the L.S. Regression The least squares fit has a number of important properties that can be derived from the first order 8 Notice also, that the rescaling of X into X * has no effect on the predicted value of sales for store #1. When advertising is measured in thousands of dollars, the predicted value of sales is x15 = 950, which represents $950,000. When advertising is measured in dollars, the predicted value of sales is x15,000 = 950, which also represents $950,000.

22 Econometrics Text by D M Prescott Chapter 3, 22 conditions [3.4] and [3.5]. In this section the following properties will be demonstrated. Least Squares Property #1 If the least squares line includes an intercept term (the line is not forced through the origin) then the sum and mean of the least squares residuals is zero, i.e.. Least Squares Property #2 The sum of the cross products between the explanatory variable X and the least squares errors is zero, i.e.,. When an intercept is included in the least squares equation, this means that Cov(X, e) = 0 and Corr(X, e) = 0. Least Squares Property #3 The sum of the cross products between the least squares errors and the predicted values of the dependent variable,, is zero, i.e.,. When an intercept is included in the least squares equation this means that Property #1 is based on equation [3.6] which was derived from the partial derivative of the sum of squared errors (first order condition [3.4]). It is reproduced here for convenience. [3.6] Recall that the least squares errors were defined in equation [4.3] to be so equation [3.6] implies that the sum of the least squares errors is zero, that is. Clearly, if the sum of the least squares errors is zero, then the average least squares error is zero as well. Another way to think of this property of the least squares fit is that the least squares line passes through the mean point of the data The mean of X in the advertising/sales example is 20 and when this is substituted into the equation of the least squares line, the result is

23 Econometrics Text by D M Prescott Chapter 3, 23. In other words, when the mean value of X is substituted into the equation of the least squares line, the result is the mean value of Y. This is not an accident due to the numbers we have chosen, it is a property of least squares that holds in every case and is directly related to the fact that the sample mean least squares error is zero. However, it is important to note that these conclusions are derived from the partial derivative of the sum of squared errors with respect to the intercept parameter a. This presupposes that least squares is free to determine the intercept parameter. If the intercept is not included (effectively, fixed at zero), then the least squares errors will generally not sum to zero and the least squares line will not pass through the sample mean. The second of the first order conditions, [3.5], is the basis of L.S. Property #2, which says that the least squares errors are uncorrelated with the explanatory variable X. The partial derivative of the sum of squared errors with respect to the slope coefficient b takes the form of equation [3.8]: It has just been pointed out that the term in parentheses is the least squares error, so [3.8] can be written as Recall that the sample covariance between two variables Z and W is

24 Econometrics Text by D M Prescott Chapter 3, 24 Clearly, if either (or both) of the means of Z and W is zero, then the covariance formula simplifies to It has already been shown that so it follows from equation [3.10] that the covariance of the least squares errors and the explanatory variable X is zero, i.e., Cov(X, e) = 0. Since the numerator of the correlation coefficient between two variables is the covariance between these same variables, it also follows that e and X are uncorrelated. Let's consider the intuition behind this property of least squares. The basic problem that least squares is trying to solve is to find the particular equation Y^ = a + bx that best explains the variable Y. The value of Y is broken down into two parts, Y = Y^ + e. The first component, Y^, is the part of Y that is explained by X - the fitted line translates changes in X into changes in predicted values of Y. The second component, e, is the error term and this is the part of Y that cannot be explained by X. But what does it mean to say that X cannot "explain" e? Suppose that X and e are positively correlated so that Cov(X, e) > 0. A scatter plot of X and e would reveal that whenever X is above its average value, e tends to be above its average value as well and when X is below average e tends to be below average. But if this is true, then increases in X would be associated with increases in e. In other words, changes in X would "explain" changes in e. This situation is clearly not consistent with the idea that the error e represents the part of Y that cannot be explained by X. To say that X cannot explain e is the same thing as saying X and e are uncorrelated and this is precisely what equation [3.11] means 9. The calculations in Table 3.9 illustrate the two important properties of least squares that have been discussed in this section. The first two columns of Table 3.9 present the original advertising and 9 In fact, if Z and W are uncorrelated we can say only that X cannot be explained by linear equations in Z (and vice versa). As shown in Chapter 2, it is possible to find examples in which Z and W are uncorrelated yet functionally related in a nonlinear way.

25 Econometrics Text by D M Prescott Chapter 3, 25 sales data. The predicted values of Y corresponding to each level of advertising expenditures are in the third column. These predicted sales levels all lie on the least squares line. The fourth column presents the differences between actual and predicted sales, i.e. the least squares errors e i. Notice that the sum of the least squares errors is zero, i.e. G e i = 0. To demonstrate that the explanatory variable X is uncorrelated with the least squares errors, the fifth column presents the products e i X i. Summing all the numbers in the fifth column shows that Ge i X i = 0. Since the mean error is zero, this implies that Cov(X, e) = 0, which in turn means that the correlation coefficient between X and e is also zero. Finally, consider L.S. Property #3 which says that the predicted values of the dependent variable are uncorrelated with the least squares errors. A numerical illustration is given in Table 3.9. The products are obtained by multiplying together the elements in columns three and four. The sum of these products, is (950)(50) + (900)(-35) (1120)(-25) = 0 The general result can be shown algebraically as follows: Notice that the two unsubscripted constants, a and b, can be factored to the front of the summation signs. Also, the two sums in the second line are both zero as direct results of L.S. Properties #1 and #2. Since the mean error is zero, the result implies that Cov(Y^, e) = 0.

26 Econometrics Text by D M Prescott Chapter 3, 26 Table 3.9 Some Properties of the Least Squares Fit (Advertising) X i (Sales) Y i X - = 20 Ȳ= 1000 (GY^i)/n = 1000 G e i = 0 G e i X i = 0 To better understand why least squares predicted values are uncorrelated with the least squares errors consider the advertising/sales example. Suppose that as the Vice President's research assistant you have calculated a linear relationship between Y and X that produces predicted values Y^ that are positively correlated with the errors, i.e. Cov(Y^, e) > 0. The VP of Sales is likely to point out that your predictions seem to have a systematic error. Stores with high advertising expenditures have high predicted sales and since Cov(Y^, e) > 0, these types of stores tend to have positive errors (sales are under predicted since actual sales lie above the fitted line). Also, stores with low advertising budgets and lower than average sales tend to below average (negative) errors, that is, sales are over predicted. Since there is a systematic relationship between the prediction errors and the level of sales, the VP will argue that when you present a sales prediction for a store that has above average advertising expenditures, she should lower your sales prediction because she knows you systematically over predict sales in such cases. However, if you present the VP of Sales with the least squares equation, you can be confident that Cov(Y^, e) = 0. The least squares predicted sales figures have errors that exhibit no systematic pattern that could be used to

27 Econometrics Text by D M Prescott Chapter 3, 27 improve the forecast 10. Finally, it should be pointed out that L.S. Property #3 actually follows from L.S. Property #1. Cov(e, Y^) = Cov(e, a + bx) = Cov(e, a) + bcov(e, X) Since a is a constant, Cov(e, a) = 0 and L.S. Property #2 states that Cov(e, X) = Measuring Goodness of Fit By definition, the least squares equation provides the best fitting line through the data. But how good is the best fitting line at explaining the observed variations in sales from store store. One way to judge how well least squares has done is to compute a statistic known as R-squared. Essentially, R- squared quantifies how useful the information on advertising is for explaining (or predicting) store sales. 11 The fundamental problem is to explain the variation in the dependent variable Y. The total variation in Y is referred to as the Total Sum of Squares, TSS, and is measured by Notice that TSS is closely related to the concept of sample variance of Y, which is TSS/n. Recall that the variance is the average value of the squared deviations of Y around its mean. TSS is the total of the squared deviations of Y around its mean. Whereas the variance does not depend on the size of the sample, clearly TSS will tend to increase with the number of observations. 10 The fitted values of Y have been referred to as predicted values of Y, but it would be better to say they are "within sample" predicted values because the actual values of Y are known to the researcher and indeed have been used to compute the "predicted" values of Y. In a real forecasting situation the forecaster does not know what the actual value of Y will be. Such forecasts go beyond the current sample and are referred to as "out of sample" predictions or forecasts. 11 One should keep in mind that it often seems straightforward to explain the past but not as easy to predict the future. R-squared measures how well one can explain the available data, but it is not a guaranteed guide to future predictive performance of the least fit.

28 Econometrics Text by D M Prescott Chapter 3, 28 An important feature of the least squares fit is that the Total Sum of Squares can be decomposed into two parts: the Regression Sum of Squares, RSS, and the Sum of Squares Residuals, SSR. The explained part of Y is The unexplained part of Y is the least squares residual, e, so SSR = G(e i ) 2. The decomposition property of least squares can be stated as TSS = RSS + SSR. Algebraically, the decomposition formula is: Proof To prove this important decomposition, begin with the left hand side and substitute Now open up the square brackets treating as two separate terms. The first two terms on the right hand side are SSR and RSS respectively, so to complete the proof it is necessary to show that the last sum is zero. Notice that on the right hand side, the first sum is zero by L.S. Property #3 and the second sum is zero by L.S. Property #1. (Notice that can be brought through the summation sign because it is an unsubscripted constant.) This completes the proof that

29 Econometrics Text by D M Prescott Chapter 3, 29 that is: TSS = SSR + RSS This decomposition of the total sum of squares provides the foundation for the goodness of fit measure known as R-squared, or R 2. Divide through by TSS and obtain 1 = RSS/TSS + SSR/TSS which shows that the proportion of the total sum of squares that is explained by the regression(rss/tss) plus the proportion that remains unexplained (SSR/TSS) add up to one. R-squared is defined as the proportion of the total sum of squares that is explained, that is, ] Interpreting R-squared First, it is straightforward to show that the goodness of fit measure R 2 always lies between 0 and 1. Since TSS, RSS and SSR are all sums of squared items, it follows that none of these sums can be negative. To better understand what R 2 measures, rewrite the decomposition of the total sum of squares in terms of

30 Econometrics Text by D M Prescott Chapter 3, 30 variances by dividing equation [3.12] throughout by n, the number of observations. The result is This result could also have been found by using the variance of a sum rule (see Chapter 2) But L.S. Property #3 says that, which implies the variance of the dependent variable is the sum of two variances. The first of these is the variance of the explained component of Y,, and the second is the variance of the least squares residuals - the unexplained component of Y. R-squared can be expressed in terms of these variances: This demonstrates that R-squared measures the proportion of the unconditional variance of Y that can be explained by the least squares fit. An interesting observation that can be drawn from equations [3.12] and [3.15] is that the least squares coefficients maximize R-squared - no other line could produce a set of predicted values of Y with a higher variance than the least squares predictions. This follows from the fact that, by definition, least squares minimizes the sum of squared residuals. Figure 3.2 illustrates the decomposition of the variance. Since the concept of variance is not easily represented graphically, the range is used to approximate the variance. The L.S. regression line translates the range of X, R(X), into the range of. That is, the minimum value of X in the sample predicts the smallest value of and similarly the maximum value of X predicts the maximum value of in the sample. Notice that since lies on the regression line, the range of is not as large as the

31 Econometrics Text by D M Prescott Chapter 3, 31 range of the observed values of Y that are dispersed above and below the regression line. This illustrates Var( Y $ ) Var( Y) the point that in all samples. What does it mean to say that X explains Y? Suppose Y is the market price of a house and X is the house size in square feet. In the housing market, prices vary from house to house and this variability can be measured by the unconditional variance of prices. It is this variance that the model seeks to explain. A regression of price on size yields least squares coefficients and a set of predicted prices that all lie on the fitted regression line. If size explains price then the regression equation should predict a wide range of prices for different sizes. Thus if the variance of the predicted prices is large and close to the variance of observed prices, then the regression equation explains a large portion of the variance of prices. In Figure 3.2, a steep regression line contributes to a high R-squared. A relatively flat regression line is associated with a low R-squared. Notice that in the extreme case that the regression line if horizontal (the least squares coefficient on X is precisely zero, R-squared is zero. Figure 3.2 can also explain why R-squared is essentially unaffected by the sample size. Note that the sample size can be increased without affecting the unconditional variance of Y, the variance of the predicted value of Y or the variance of X. Figure 3.2 remains unchanged except that more and more data are packed into the parallelogram around the regression line. The quantity or density of points in this parallelogram has no bearing on R-squared - what matters is the relationship between the variances. In short, simply increasing the sample size will not help to increase the proportion of the variation in Y that can be explained by X. Finally, the fact that the name R-squared has the term "squared" in it raises the question of what R = %(R 2 ) represents. It turns out the R-squared is the square of the correlation coefficient between Y and so R = Corr(Y, Y^). It makes intuitive sense that the closer the fitted values, are to Y, the higher will be the R-squared statistic. The proof of this is straightforward.

ECON3150/4150 Spring 2015

ECON3150/4150 Spring 2015 ECON3150/4150 Spring 2015 Lecture 3&4 - The linear regression model Siv-Elisabeth Skjelbred University of Oslo January 29, 2015 1 / 67 Chapter 4 in S&W Section 17.1 in S&W (extended OLS assumptions) 2

More information

Chapter 7. Testing Linear Restrictions on Regression Coefficients

Chapter 7. Testing Linear Restrictions on Regression Coefficients Chapter 7 Testing Linear Restrictions on Regression Coefficients 1.F-tests versus t-tests In the previous chapter we discussed several applications of the t-distribution to testing hypotheses in the linear

More information

Business Economics BUSINESS ECONOMICS. PAPER No. : 8, FUNDAMENTALS OF ECONOMETRICS MODULE No. : 3, GAUSS MARKOV THEOREM

Business Economics BUSINESS ECONOMICS. PAPER No. : 8, FUNDAMENTALS OF ECONOMETRICS MODULE No. : 3, GAUSS MARKOV THEOREM Subject Business Economics Paper No and Title Module No and Title Module Tag 8, Fundamentals of Econometrics 3, The gauss Markov theorem BSE_P8_M3 1 TABLE OF CONTENTS 1. INTRODUCTION 2. ASSUMPTIONS OF

More information

Applied Regression Modeling: A Business Approach Chapter 2: Simple Linear Regression Sections

Applied Regression Modeling: A Business Approach Chapter 2: Simple Linear Regression Sections Applied Regression Modeling: A Business Approach Chapter 2: Simple Linear Regression Sections 2.1 2.3 by Iain Pardoe 2.1 Probability model for and 2 Simple linear regression model for and....................................

More information

Wooldridge, Introductory Econometrics, 4th ed. Chapter 2: The simple regression model

Wooldridge, Introductory Econometrics, 4th ed. Chapter 2: The simple regression model Wooldridge, Introductory Econometrics, 4th ed. Chapter 2: The simple regression model Most of this course will be concerned with use of a regression model: a structure in which one or more explanatory

More information

ECON2228 Notes 2. Christopher F Baum. Boston College Economics. cfb (BC Econ) ECON2228 Notes / 47

ECON2228 Notes 2. Christopher F Baum. Boston College Economics. cfb (BC Econ) ECON2228 Notes / 47 ECON2228 Notes 2 Christopher F Baum Boston College Economics 2014 2015 cfb (BC Econ) ECON2228 Notes 2 2014 2015 1 / 47 Chapter 2: The simple regression model Most of this course will be concerned with

More information

Section 3: Simple Linear Regression

Section 3: Simple Linear Regression Section 3: Simple Linear Regression Carlos M. Carvalho The University of Texas at Austin McCombs School of Business http://faculty.mccombs.utexas.edu/carlos.carvalho/teaching/ 1 Regression: General Introduction

More information

Mathematics for Economics MA course

Mathematics for Economics MA course Mathematics for Economics MA course Simple Linear Regression Dr. Seetha Bandara Simple Regression Simple linear regression is a statistical method that allows us to summarize and study relationships between

More information

Lectures 5 & 6: Hypothesis Testing

Lectures 5 & 6: Hypothesis Testing Lectures 5 & 6: Hypothesis Testing in which you learn to apply the concept of statistical significance to OLS estimates, learn the concept of t values, how to use them in regression work and come across

More information

Simple Linear Regression

Simple Linear Regression Simple Linear Regression ST 430/514 Recall: A regression model describes how a dependent variable (or response) Y is affected, on average, by one or more independent variables (or factors, or covariates)

More information

regression analysis is a type of inferential statistics which tells us whether relationships between two or more variables exist

regression analysis is a type of inferential statistics which tells us whether relationships between two or more variables exist regression analysis is a type of inferential statistics which tells us whether relationships between two or more variables exist sales $ (y - dependent variable) advertising $ (x - independent variable)

More information

Chapter 2: simple regression model

Chapter 2: simple regression model Chapter 2: simple regression model Goal: understand how to estimate and more importantly interpret the simple regression Reading: chapter 2 of the textbook Advice: this chapter is foundation of econometrics.

More information

Chapter 16. Simple Linear Regression and dcorrelation

Chapter 16. Simple Linear Regression and dcorrelation Chapter 16 Simple Linear Regression and dcorrelation 16.1 Regression Analysis Our problem objective is to analyze the relationship between interval variables; regression analysis is the first tool we will

More information

AP Statistics Bivariate Data Analysis Test Review. Multiple-Choice

AP Statistics Bivariate Data Analysis Test Review. Multiple-Choice Name Period AP Statistics Bivariate Data Analysis Test Review Multiple-Choice 1. The correlation coefficient measures: (a) Whether there is a relationship between two variables (b) The strength of the

More information

ECON The Simple Regression Model

ECON The Simple Regression Model ECON 351 - The Simple Regression Model Maggie Jones 1 / 41 The Simple Regression Model Our starting point will be the simple regression model where we look at the relationship between two variables In

More information

Regression Analysis. BUS 735: Business Decision Making and Research

Regression Analysis. BUS 735: Business Decision Making and Research Regression Analysis BUS 735: Business Decision Making and Research 1 Goals and Agenda Goals of this section Specific goals Learn how to detect relationships between ordinal and categorical variables. Learn

More information

The Simple Linear Regression Model

The Simple Linear Regression Model The Simple Linear Regression Model Lesson 3 Ryan Safner 1 1 Department of Economics Hood College ECON 480 - Econometrics Fall 2017 Ryan Safner (Hood College) ECON 480 - Lesson 3 Fall 2017 1 / 77 Bivariate

More information

Chapter 4: Regression Models

Chapter 4: Regression Models Sales volume of company 1 Textbook: pp. 129-164 Chapter 4: Regression Models Money spent on advertising 2 Learning Objectives After completing this chapter, students will be able to: Identify variables,

More information

Regression Analysis. Ordinary Least Squares. The Linear Model

Regression Analysis. Ordinary Least Squares. The Linear Model Regression Analysis Linear regression is one of the most widely used tools in statistics. Suppose we were jobless college students interested in finding out how big (or small) our salaries would be 20

More information

ECON 497 Midterm Spring

ECON 497 Midterm Spring ECON 497 Midterm Spring 2009 1 ECON 497: Economic Research and Forecasting Name: Spring 2009 Bellas Midterm You have three hours and twenty minutes to complete this exam. Answer all questions and explain

More information

Chapter 16. Simple Linear Regression and Correlation

Chapter 16. Simple Linear Regression and Correlation Chapter 16 Simple Linear Regression and Correlation 16.1 Regression Analysis Our problem objective is to analyze the relationship between interval variables; regression analysis is the first tool we will

More information

Algebra II Notes Quadratic Functions Unit Applying Quadratic Functions. Math Background

Algebra II Notes Quadratic Functions Unit Applying Quadratic Functions. Math Background Applying Quadratic Functions Math Background Previously, you Graphed and solved quadratic functions. Solved literal equations for a given variable. Found the inverse for a linear function. Verified by

More information

1 A Non-technical Introduction to Regression

1 A Non-technical Introduction to Regression 1 A Non-technical Introduction to Regression Chapters 1 and Chapter 2 of the textbook are reviews of material you should know from your previous study (e.g. in your second year course). They cover, in

More information

Regression Analysis. BUS 735: Business Decision Making and Research. Learn how to detect relationships between ordinal and categorical variables.

Regression Analysis. BUS 735: Business Decision Making and Research. Learn how to detect relationships between ordinal and categorical variables. Regression Analysis BUS 735: Business Decision Making and Research 1 Goals of this section Specific goals Learn how to detect relationships between ordinal and categorical variables. Learn how to estimate

More information

APPENDIX 1 BASIC STATISTICS. Summarizing Data

APPENDIX 1 BASIC STATISTICS. Summarizing Data 1 APPENDIX 1 Figure A1.1: Normal Distribution BASIC STATISTICS The problem that we face in financial analysis today is not having too little information but too much. Making sense of large and often contradictory

More information

Regression Models. Chapter 4. Introduction. Introduction. Introduction

Regression Models. Chapter 4. Introduction. Introduction. Introduction Chapter 4 Regression Models Quantitative Analysis for Management, Tenth Edition, by Render, Stair, and Hanna 008 Prentice-Hall, Inc. Introduction Regression analysis is a very valuable tool for a manager

More information

2. Linear regression with multiple regressors

2. Linear regression with multiple regressors 2. Linear regression with multiple regressors Aim of this section: Introduction of the multiple regression model OLS estimation in multiple regression Measures-of-fit in multiple regression Assumptions

More information

ECON3150/4150 Spring 2016

ECON3150/4150 Spring 2016 ECON3150/4150 Spring 2016 Lecture 4 - The linear regression model Siv-Elisabeth Skjelbred University of Oslo Last updated: January 26, 2016 1 / 49 Overview These lecture slides covers: The linear regression

More information

LI EAR REGRESSIO A D CORRELATIO

LI EAR REGRESSIO A D CORRELATIO CHAPTER 6 LI EAR REGRESSIO A D CORRELATIO Page Contents 6.1 Introduction 10 6. Curve Fitting 10 6.3 Fitting a Simple Linear Regression Line 103 6.4 Linear Correlation Analysis 107 6.5 Spearman s Rank Correlation

More information

Ch 13 & 14 - Regression Analysis

Ch 13 & 14 - Regression Analysis Ch 3 & 4 - Regression Analysis Simple Regression Model I. Multiple Choice:. A simple regression is a regression model that contains a. only one independent variable b. only one dependent variable c. more

More information

Calculus. Applications of Differentiations (IV)

Calculus. Applications of Differentiations (IV) Calculus Applications of Differentiations (IV) Outline 1 Rates of Change In Economics And The Sciences Applications of Derivative In Economics Applications of Derivative in the Sciences 2 Related Rate

More information

Relationships Between Quantities

Relationships Between Quantities Algebra 1 Relationships Between Quantities Relationships Between Quantities Everyone loves math until there are letters (known as variables) in problems!! Do students complain about reading when they come

More information

Midterm 2 - Solutions

Midterm 2 - Solutions Ecn 102 - Analysis of Economic Data University of California - Davis February 24, 2010 Instructor: John Parman Midterm 2 - Solutions You have until 10:20am to complete this exam. Please remember to put

More information

11.5 Regression Linear Relationships

11.5 Regression Linear Relationships Contents 11.5 Regression............................. 835 11.5.1 Linear Relationships................... 835 11.5.2 The Least Squares Regression Line........... 837 11.5.3 Using the Regression Line................

More information

SCHOOL OF DISTANCE EDUCATION

SCHOOL OF DISTANCE EDUCATION SCHOOL OF DISTANCE EDUCATION CCSS UG PROGRAMME MATHEMATICS (OPEN COURSE) (For students not having Mathematics as Core Course) MM5D03: MATHEMATICS FOR SOCIAL SCIENCES FIFTH SEMESTER STUDY NOTES Prepared

More information

Biostatistics and Design of Experiments Prof. Mukesh Doble Department of Biotechnology Indian Institute of Technology, Madras

Biostatistics and Design of Experiments Prof. Mukesh Doble Department of Biotechnology Indian Institute of Technology, Madras Biostatistics and Design of Experiments Prof. Mukesh Doble Department of Biotechnology Indian Institute of Technology, Madras Lecture - 39 Regression Analysis Hello and welcome to the course on Biostatistics

More information

CLASS NOTES: BUSINESS CALCULUS

CLASS NOTES: BUSINESS CALCULUS CLASS NOTES: BUSINESS CALCULUS These notes can be thought of as the logical skeleton of my lectures, although they will generally contain a fuller exposition of concepts but fewer examples than my lectures.

More information

Ordinary Least Squares Regression Explained: Vartanian

Ordinary Least Squares Regression Explained: Vartanian Ordinary Least Squares Regression Explained: Vartanian When to Use Ordinary Least Squares Regression Analysis A. Variable types. When you have an interval/ratio scale dependent variable.. When your independent

More information

Glossary. The ISI glossary of statistical terms provides definitions in a number of different languages:

Glossary. The ISI glossary of statistical terms provides definitions in a number of different languages: Glossary The ISI glossary of statistical terms provides definitions in a number of different languages: http://isi.cbs.nl/glossary/index.htm Adjusted r 2 Adjusted R squared measures the proportion of the

More information

Probability Distributions

Probability Distributions CONDENSED LESSON 13.1 Probability Distributions In this lesson, you Sketch the graph of the probability distribution for a continuous random variable Find probabilities by finding or approximating areas

More information

ECON3150/4150 Spring 2016

ECON3150/4150 Spring 2016 ECON3150/4150 Spring 2016 Lecture 6 Multiple regression model Siv-Elisabeth Skjelbred University of Oslo February 5th Last updated: February 3, 2016 1 / 49 Outline Multiple linear regression model and

More information

Fitting a regression model

Fitting a regression model Fitting a regression model We wish to fit a simple linear regression model: y = β 0 + β 1 x + ɛ. Fitting a model means obtaining estimators for the unknown population parameters β 0 and β 1 (and also for

More information

Industrial Engineering Prof. Inderdeep Singh Department of Mechanical & Industrial Engineering Indian Institute of Technology, Roorkee

Industrial Engineering Prof. Inderdeep Singh Department of Mechanical & Industrial Engineering Indian Institute of Technology, Roorkee Industrial Engineering Prof. Inderdeep Singh Department of Mechanical & Industrial Engineering Indian Institute of Technology, Roorkee Module - 04 Lecture - 05 Sales Forecasting - II A very warm welcome

More information

An Introduction to Parameter Estimation

An Introduction to Parameter Estimation Introduction Introduction to Econometrics An Introduction to Parameter Estimation This document combines several important econometric foundations and corresponds to other documents such as the Introduction

More information

3. Find the slope of the tangent line to the curve given by 3x y e x+y = 1 + ln x at (1, 1).

3. Find the slope of the tangent line to the curve given by 3x y e x+y = 1 + ln x at (1, 1). 1. Find the derivative of each of the following: (a) f(x) = 3 2x 1 (b) f(x) = log 4 (x 2 x) 2. Find the slope of the tangent line to f(x) = ln 2 ln x at x = e. 3. Find the slope of the tangent line to

More information

Math 016 Lessons Wimayra LUY

Math 016 Lessons Wimayra LUY Math 016 Lessons Wimayra LUY wluy@ccp.edu MATH 016 Lessons LESSON 1 Natural Numbers The set of natural numbers is given by N = {0, 1, 2, 3, 4...}. Natural numbers are used for two main reasons: 1. counting,

More information

CORE. Chapter 3: Interacting Linear Functions, Linear Systems. Algebra Assessments

CORE. Chapter 3: Interacting Linear Functions, Linear Systems. Algebra Assessments CORE Algebra Assessments Chapter 3: Interacting Linear Functions, Linear Systems 97 98 Bears Band Booster Club The Bears Band Booster Club has decided to sell calendars to the band members and their parents.

More information

download instant at

download instant at Answers to Odd-Numbered Exercises Chapter One: An Overview of Regression Analysis 1-3. (a) Positive, (b) negative, (c) positive, (d) negative, (e) ambiguous, (f) negative. 1-5. (a) The coefficients in

More information

Mathematics: applications and interpretation SL

Mathematics: applications and interpretation SL Mathematics: applications and interpretation SL Chapter 1: Approximations and error A Rounding numbers B Approximations C Errors in measurement D Absolute and percentage error The first two sections of

More information

ACCESS TO SCIENCE, ENGINEERING AND AGRICULTURE: MATHEMATICS 1 MATH00030 SEMESTER /2018

ACCESS TO SCIENCE, ENGINEERING AND AGRICULTURE: MATHEMATICS 1 MATH00030 SEMESTER /2018 ACCESS TO SCIENCE, ENGINEERING AND AGRICULTURE: MATHEMATICS 1 MATH00030 SEMESTER 1 2017/2018 DR. ANTHONY BROWN 1. Arithmetic and Algebra 1.1. Arithmetic of Numbers. While we have calculators and computers

More information

9 Correlation and Regression

9 Correlation and Regression 9 Correlation and Regression SW, Chapter 12. Suppose we select n = 10 persons from the population of college seniors who plan to take the MCAT exam. Each takes the test, is coached, and then retakes the

More information

Chapter 9. Correlation and Regression

Chapter 9. Correlation and Regression Chapter 9 Correlation and Regression Lesson 9-1/9-2, Part 1 Correlation Registered Florida Pleasure Crafts and Watercraft Related Manatee Deaths 100 80 60 40 20 0 1991 1993 1995 1997 1999 Year Boats in

More information

ABE Math Review Package

ABE Math Review Package P a g e ABE Math Review Package This material is intended as a review of skills you once learned and wish to review before your assessment. Before studying Algebra, you should be familiar with all of the

More information

Chapter 13 - Inverse Functions

Chapter 13 - Inverse Functions Chapter 13 - Inverse Functions In the second part of this book on Calculus, we shall be devoting our study to another type of function, the exponential function and its close relative the Sine function.

More information

1 A Review of Correlation and Regression

1 A Review of Correlation and Regression 1 A Review of Correlation and Regression SW, Chapter 12 Suppose we select n = 10 persons from the population of college seniors who plan to take the MCAT exam. Each takes the test, is coached, and then

More information

In the previous chapter, we learned how to use the method of least-squares

In the previous chapter, we learned how to use the method of least-squares 03-Kahane-45364.qxd 11/9/2007 4:40 PM Page 37 3 Model Performance and Evaluation In the previous chapter, we learned how to use the method of least-squares to find a line that best fits a scatter of points.

More information

Regression Models REVISED TEACHING SUGGESTIONS ALTERNATIVE EXAMPLES

Regression Models REVISED TEACHING SUGGESTIONS ALTERNATIVE EXAMPLES M04_REND6289_10_IM_C04.QXD 5/7/08 2:49 PM Page 46 4 C H A P T E R Regression Models TEACHING SUGGESTIONS Teaching Suggestion 4.1: Which Is the Independent Variable? We find that students are often confused

More information

Massachusetts Tests for Educator Licensure (MTEL )

Massachusetts Tests for Educator Licensure (MTEL ) Massachusetts Tests for Educator Licensure (MTEL ) BOOKLET 2 Mathematics Subtest Copyright 2010 Pearson Education, Inc. or its affiliate(s). All rights reserved. Evaluation Systems, Pearson, P.O. Box 226,

More information

Quarter 2 400, , , , , , ,000 50,000

Quarter 2 400, , , , , , ,000 50,000 Algebra 2 Quarter 2 Quadratic Functions Introduction to Polynomial Functions Hybrid Electric Vehicles Since 1999, there has been a growing trend in the sales of hybrid electric vehicles. These data show

More information

Keller: Stats for Mgmt & Econ, 7th Ed July 17, 2006

Keller: Stats for Mgmt & Econ, 7th Ed July 17, 2006 Chapter 17 Simple Linear Regression and Correlation 17.1 Regression Analysis Our problem objective is to analyze the relationship between interval variables; regression analysis is the first tool we will

More information

MATH 1130 Exam 1 Review Sheet

MATH 1130 Exam 1 Review Sheet MATH 1130 Exam 1 Review Sheet The Cartesian Coordinate Plane The Cartesian Coordinate Plane is a visual representation of the collection of all ordered pairs (x, y) where x and y are real numbers. This

More information

MATH 080 Final-Exam Review

MATH 080 Final-Exam Review MATH 080 Final-Exam Review Can you simplify an expression using the order of operations? 1) Simplify 32(11-8) - 18 3 2-3 2) Simplify 5-3 3-3 6 + 3 A) 5 9 B) 19 9 C) - 25 9 D) 25 9 Can you evaluate an algebraic

More information

Algebra 31 Summer Work Packet Review and Study Guide

Algebra 31 Summer Work Packet Review and Study Guide Algebra Summer Work Packet Review and Study Guide This study guide is designed to accompany the Algebra Summer Work Packet. Its purpose is to offer a review of the ten specific concepts covered in the

More information

1 Correlation and Inference from Regression

1 Correlation and Inference from Regression 1 Correlation and Inference from Regression Reading: Kennedy (1998) A Guide to Econometrics, Chapters 4 and 6 Maddala, G.S. (1992) Introduction to Econometrics p. 170-177 Moore and McCabe, chapter 12 is

More information

The Simple Regression Model. Simple Regression Model 1

The Simple Regression Model. Simple Regression Model 1 The Simple Regression Model Simple Regression Model 1 Simple regression model: Objectives Given the model: - where y is earnings and x years of education - Or y is sales and x is spending in advertising

More information

Business Statistics. Lecture 9: Simple Regression

Business Statistics. Lecture 9: Simple Regression Business Statistics Lecture 9: Simple Regression 1 On to Model Building! Up to now, class was about descriptive and inferential statistics Numerical and graphical summaries of data Confidence intervals

More information

The Not-Formula Book for C2 Everything you need to know for Core 2 that won t be in the formula book Examination Board: AQA

The Not-Formula Book for C2 Everything you need to know for Core 2 that won t be in the formula book Examination Board: AQA Not The Not-Formula Book for C Everything you need to know for Core that won t be in the formula book Examination Board: AQA Brief This document is intended as an aid for revision. Although it includes

More information

GRE Quantitative Reasoning Practice Questions

GRE Quantitative Reasoning Practice Questions GRE Quantitative Reasoning Practice Questions y O x 7. The figure above shows the graph of the function f in the xy-plane. What is the value of f (f( ))? A B C 0 D E Explanation Note that to find f (f(

More information

Lecture Notes. Applied Mathematics for Business, Economics, and the Social Sciences (4th Edition); by Frank S. Budnick

Lecture Notes. Applied Mathematics for Business, Economics, and the Social Sciences (4th Edition); by Frank S. Budnick 1 Lecture Notes Applied Mathematics for Business, Economics, and the Social Sciences (4th Edition); by Frank S. Budnick 2 Chapter 2: Linear Equations Definition: Linear equations are first degree equations.

More information

Applied Regression Modeling: A Business Approach Chapter 3: Multiple Linear Regression Sections

Applied Regression Modeling: A Business Approach Chapter 3: Multiple Linear Regression Sections Applied Regression Modeling: A Business Approach Chapter 3: Multiple Linear Regression Sections 3.1 3.3.2 by Iain Pardoe 3.1 Probability model for (X 1, X 2,...) and Y 2 Multiple linear regression................................................

More information

Year 10 Mathematics Semester 2 Bivariate Data Chapter 13

Year 10 Mathematics Semester 2 Bivariate Data Chapter 13 Year 10 Mathematics Semester 2 Bivariate Data Chapter 13 Why learn this? Observations of two or more variables are often recorded, for example, the heights and weights of individuals. Studying the data

More information

Wooldridge, Introductory Econometrics, 4th ed. Chapter 6: Multiple regression analysis: Further issues

Wooldridge, Introductory Econometrics, 4th ed. Chapter 6: Multiple regression analysis: Further issues Wooldridge, Introductory Econometrics, 4th ed. Chapter 6: Multiple regression analysis: Further issues What effects will the scale of the X and y variables have upon multiple regression? The coefficients

More information

V. Properties of estimators {Parts C, D & E in this file}

V. Properties of estimators {Parts C, D & E in this file} A. Definitions & Desiderata. model. estimator V. Properties of estimators {Parts C, D & E in this file}. sampling errors and sampling distribution 4. unbiasedness 5. low sampling variance 6. low mean squared

More information

EC212: Introduction to Econometrics Review Materials (Wooldridge, Appendix)

EC212: Introduction to Econometrics Review Materials (Wooldridge, Appendix) 1 EC212: Introduction to Econometrics Review Materials (Wooldridge, Appendix) Taisuke Otsu London School of Economics Summer 2018 A.1. Summation operator (Wooldridge, App. A.1) 2 3 Summation operator For

More information

Econometric Modelling Prof. Rudra P. Pradhan Department of Management Indian Institute of Technology, Kharagpur

Econometric Modelling Prof. Rudra P. Pradhan Department of Management Indian Institute of Technology, Kharagpur Econometric Modelling Prof. Rudra P. Pradhan Department of Management Indian Institute of Technology, Kharagpur Module No. # 01 Lecture No. # 28 LOGIT and PROBIT Model Good afternoon, this is doctor Pradhan

More information

Linear Functions, Equations, and Inequalities

Linear Functions, Equations, and Inequalities CHAPTER Linear Functions, Equations, and Inequalities Inventory is the list of items that businesses stock in stores and warehouses to supply customers. Businesses in the United States keep about.5 trillion

More information

MBF1923 Econometrics Prepared by Dr Khairul Anuar

MBF1923 Econometrics Prepared by Dr Khairul Anuar MBF1923 Econometrics Prepared by Dr Khairul Anuar L4 Ordinary Least Squares www.notes638.wordpress.com Ordinary Least Squares The bread and butter of regression analysis is the estimation of the coefficient

More information

CHAPTER 8 INTRODUCTION TO STATISTICAL ANALYSIS

CHAPTER 8 INTRODUCTION TO STATISTICAL ANALYSIS CHAPTER 8 INTRODUCTION TO STATISTICAL ANALYSIS LEARNING OBJECTIVES: After studying this chapter, a student should understand: notation used in statistics; how to represent variables in a mathematical form

More information

Lecture 3: Multiple Regression

Lecture 3: Multiple Regression Lecture 3: Multiple Regression R.G. Pierse 1 The General Linear Model Suppose that we have k explanatory variables Y i = β 1 + β X i + β 3 X 3i + + β k X ki + u i, i = 1,, n (1.1) or Y i = β j X ji + u

More information

Simple Linear Regression Estimation and Properties

Simple Linear Regression Estimation and Properties Simple Linear Regression Estimation and Properties Outline Review of the Reading Estimate parameters using OLS Other features of OLS Numerical Properties of OLS Assumptions of OLS Goodness of Fit Checking

More information

9. Linear Regression and Correlation

9. Linear Regression and Correlation 9. Linear Regression and Correlation Data: y a quantitative response variable x a quantitative explanatory variable (Chap. 8: Recall that both variables were categorical) For example, y = annual income,

More information

Chapter 4. Systems of Linear Equations; Matrices. Opening Example. Section 1 Review: Systems of Linear Equations in Two Variables

Chapter 4. Systems of Linear Equations; Matrices. Opening Example. Section 1 Review: Systems of Linear Equations in Two Variables Chapter 4 Systems of Linear Equations; Matrices Section 1 Review: Systems of Linear Equations in Two Variables Opening Example A restaurant serves two types of fish dinners- small for $5.99 and large for

More information

Finite Mathematics : A Business Approach

Finite Mathematics : A Business Approach Finite Mathematics : A Business Approach Dr. Brian Travers and Prof. James Lampes Second Edition Cover Art by Stephanie Oxenford Additional Editing by John Gambino Contents What You Should Already Know

More information

SHORT ANSWER. Write the word or phrase that best completes each statement or answers the question. x )

SHORT ANSWER. Write the word or phrase that best completes each statement or answers the question. x ) Midterm Review Name SHORT ANSWER. Write the word or phrase that best completes each statement or answers the question. Decide whether or not the arrow diagram defines a function. 1) Domain Range 1) Determine

More information

Business Statistics. Chapter 14 Introduction to Linear Regression and Correlation Analysis QMIS 220. Dr. Mohammad Zainal

Business Statistics. Chapter 14 Introduction to Linear Regression and Correlation Analysis QMIS 220. Dr. Mohammad Zainal Department of Quantitative Methods & Information Systems Business Statistics Chapter 14 Introduction to Linear Regression and Correlation Analysis QMIS 220 Dr. Mohammad Zainal Chapter Goals After completing

More information

Algebra 2 Summer Work Packet Review and Study Guide

Algebra 2 Summer Work Packet Review and Study Guide Algebra Summer Work Packet Review and Study Guide This study guide is designed to accompany the Algebra Summer Work Packet. Its purpose is to offer a review of the nine specific concepts covered in the

More information

Chapter 7: Simple linear regression

Chapter 7: Simple linear regression The absolute movement of the ground and buildings during an earthquake is small even in major earthquakes. The damage that a building suffers depends not upon its displacement, but upon the acceleration.

More information

SCHOOL OF MATHEMATICS MATHEMATICS FOR PART I ENGINEERING. Self-paced Course

SCHOOL OF MATHEMATICS MATHEMATICS FOR PART I ENGINEERING. Self-paced Course SCHOOL OF MATHEMATICS MATHEMATICS FOR PART I ENGINEERING Self-paced Course MODULE ALGEBRA Module Topics Simplifying expressions and algebraic functions Rearranging formulae Indices 4 Rationalising a denominator

More information

Math, Stats, and Mathstats Review ECONOMETRICS (ECON 360) BEN VAN KAMMEN, PHD

Math, Stats, and Mathstats Review ECONOMETRICS (ECON 360) BEN VAN KAMMEN, PHD Math, Stats, and Mathstats Review ECONOMETRICS (ECON 360) BEN VAN KAMMEN, PHD Outline These preliminaries serve to signal to students what tools they need to know to succeed in ECON 360 and refresh their

More information

UNIT 3: MODELING AND ANALYZING QUADRATIC FUNCTIONS

UNIT 3: MODELING AND ANALYZING QUADRATIC FUNCTIONS UNIT 3: MODELING AND ANALYZING QUADRATIC FUNCTIONS This unit investigates quadratic functions. Students study the structure of quadratic expressions and write quadratic expressions in equivalent forms.

More information

Regression. ECO 312 Fall 2013 Chris Sims. January 12, 2014

Regression. ECO 312 Fall 2013 Chris Sims. January 12, 2014 ECO 312 Fall 2013 Chris Sims Regression January 12, 2014 c 2014 by Christopher A. Sims. This document is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License What

More information

EC4051 Project and Introductory Econometrics

EC4051 Project and Introductory Econometrics EC4051 Project and Introductory Econometrics Dudley Cooke Trinity College Dublin Dudley Cooke (Trinity College Dublin) Intro to Econometrics 1 / 23 Project Guidelines Each student is required to undertake

More information

Section 4: Math Test Calculator

Section 4: Math Test Calculator QUESTION 0. The correct answer is 3 _ or.6. Triangle ABC is a right triangle with its right 5 angle at B. Thus, _ AC is the hypotenuse of right triangle ABC, and _ AB and _ BC are the legs of right triangle

More information

MAC 2233, Survey of Calculus, Exam 3 Review This exam covers lectures 21 29,

MAC 2233, Survey of Calculus, Exam 3 Review This exam covers lectures 21 29, MAC 2233, Survey of Calculus, Exam 3 Review This exam covers lectures 21 29, This review includes typical exam problems. It is not designed to be comprehensive, but to be representative of topics covered

More information

STA441: Spring Multiple Regression. This slide show is a free open source document. See the last slide for copyright information.

STA441: Spring Multiple Regression. This slide show is a free open source document. See the last slide for copyright information. STA441: Spring 2018 Multiple Regression This slide show is a free open source document. See the last slide for copyright information. 1 Least Squares Plane 2 Statistical MODEL There are p-1 explanatory

More information

SAMPLE. The SSAT Course Book MIDDLE & UPPER LEVEL QUANTITATIVE. Focusing on the Individual Student

SAMPLE. The SSAT Course Book MIDDLE & UPPER LEVEL QUANTITATIVE. Focusing on the Individual Student The SSAT Course Book MIDDLE & UPPER LEVEL QUANTITATIVE Focusing on the Individual Student Copyright Statement The SSAT Course Book, along with all Summit Educational Group Course Materials, is protected

More information

Geometry 21 Summer Work Packet Review and Study Guide

Geometry 21 Summer Work Packet Review and Study Guide Geometry Summer Work Packet Review and Study Guide This study guide is designed to accompany the Geometry Summer Work Packet. Its purpose is to offer a review of the ten specific concepts covered in the

More information

Chapter 6. Logistic Regression. 6.1 A linear model for the log odds

Chapter 6. Logistic Regression. 6.1 A linear model for the log odds Chapter 6 Logistic Regression In logistic regression, there is a categorical response variables, often coded 1=Yes and 0=No. Many important phenomena fit this framework. The patient survives the operation,

More information

Estimating σ 2. We can do simple prediction of Y and estimation of the mean of Y at any value of X.

Estimating σ 2. We can do simple prediction of Y and estimation of the mean of Y at any value of X. Estimating σ 2 We can do simple prediction of Y and estimation of the mean of Y at any value of X. To perform inferences about our regression line, we must estimate σ 2, the variance of the error term.

More information

Chapter 14 Student Lecture Notes 14-1

Chapter 14 Student Lecture Notes 14-1 Chapter 14 Student Lecture Notes 14-1 Business Statistics: A Decision-Making Approach 6 th Edition Chapter 14 Multiple Regression Analysis and Model Building Chap 14-1 Chapter Goals After completing this

More information