(It's not always good, but we can always make it.) (4) Convert the normal distribution N to the standard normal distribution Z. Specically.

Similar documents
Suppose that you have three coins. Coin A is fair, coin B shows heads with probability 0.6 and coin C shows heads with probability 0.8.

Counting principles, including permutations and combinations.

A Probability Primer. A random walk down a probabilistic path leading to some stochastic thoughts on chance events and uncertain outcomes.

Poisson approximations

Math Bootcamp 2012 Miscellaneous

success and failure independent from one trial to the next?

18.440: Lecture 19 Normal random variables

1 Normal Distribution.

Chapter 18 Sampling Distribution Models

Bell-shaped curves, variance

Introduction to Probability

MA 1125 Lecture 15 - The Standard Normal Distribution. Friday, October 6, Objectives: Introduce the standard normal distribution and table.

EXAM. Exam #1. Math 3342 Summer II, July 21, 2000 ANSWERS

3 Multiple Discrete Random Variables

GEOMETRIC -discrete A discrete random variable R counts number of times needed before an event occurs

Deep Learning for Computer Vision

6.041/6.431 Fall 2010 Quiz 2 Solutions

STAT 516 Midterm Exam 3 Friday, April 18, 2008

Continuous Expectation and Variance, the Law of Large Numbers, and the Central Limit Theorem Spring 2014

ACCESS TO SCIENCE, ENGINEERING AND AGRICULTURE: MATHEMATICS 2 MATH00040 SEMESTER / Probability

Topic 3: The Expectation of a Random Variable

Paper2Practice [303 marks]

More on Distribution Function

Evaluating Classifiers. Lecture 2 Instructor: Max Welling

Discrete Probability distribution Discrete Probability distribution

Lesson 100: The Normal Distribution. HL Math - Santowski

Example continued. Math 425 Intro to Probability Lecture 37. Example continued. Example

18.440: Lecture 26 Conditional expectation

1 Probability Distributions

STAT 516 Midterm Exam 2 Friday, March 7, 2008

CSE 103 Homework 8: Solutions November 30, var(x) = np(1 p) = P r( X ) 0.95 P r( X ) 0.

Lecture 8: Continuous random variables, expectation and variance

Lecture 8 Sampling Theory

PROBABILITY DISTRIBUTION

SDS 321: Introduction to Probability and Statistics

Random Variables. Saravanan Vijayakumaran Department of Electrical Engineering Indian Institute of Technology Bombay

MATH Notebook 5 Fall 2018/2019

Statistics, Probability Distributions & Error Propagation. James R. Graham

Conditional distributions (discrete case)

Biostatistics in Dentistry

CMPT 882 Machine Learning

Solving Equations by Factoring. Solve the quadratic equation x 2 16 by factoring. We write the equation in standard form: x

STAT2201. Analysis of Engineering & Scientific Data. Unit 3

Notes 12 Autumn 2005

S n = x + X 1 + X X n.

Stationary independent increments. 1. Random changes of the form X t+h X t fixed h > 0 are called increments of the process.

Advanced Herd Management Probabilities and distributions

Central Limit Theorem and the Law of Large Numbers Class 6, Jeremy Orloff and Jonathan Bloom

To find the median, find the 40 th quartile and the 70 th quartile (which are easily found at y=1 and y=2, respectively). Then we interpolate:

Find the component form of with initial point A(1, 3) and terminal point B(1, 3). Component form = 1 1, 3 ( 3) (x 1., y 1. ) = (1, 3) = 0, 6 Subtract.

Binomial random variable

Some Continuous Probability Distributions: Part I. Continuous Uniform distribution Normal Distribution. Exponential Distribution

ECO227: Term Test 2 (Solutions and Marking Procedure)

Probability and Statistics Concepts

Probability and Probability Distributions. Dr. Mohammed Alahmed

Massachusetts Institute of Technology

UNIT NUMBER PROBABILITY 6 (Statistics for the binomial distribution) A.J.Hobson

Bernoulli Trials, Binomial and Cumulative Distributions

THE ROYAL STATISTICAL SOCIETY 2007 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE (MODULAR FORMAT) MODULE 2 PROBABILITY MODELS

Discrete Probability. Chemistry & Physics. Medicine

The Central Limit Theorem

Introductory Probability

UC Berkeley Department of Electrical Engineering and Computer Science. EE 126: Probablity and Random Processes. Problem Set 9 Fall 2007

Discrete Random Variables

Mathematical Methods for Computer Science

Ch. 5 Joint Probability Distributions and Random Samples

CME 106: Review Probability theory

Special distributions

Probability Theory. Introduction to Probability Theory. Principles of Counting Examples. Principles of Counting. Probability spaces.

Quick Tour of Basic Probability Theory and Linear Algebra

Chapter 7: Theoretical Probability Distributions Variable - Measured/Categorized characteristic

INF FALL NATURAL LANGUAGE PROCESSING. Jan Tore Lønning

CDA6530: Performance Models of Computers and Networks. Chapter 2: Review of Practical Random Variables

[Read Ch. 5] [Recommended exercises: 5.2, 5.3, 5.4]

Lecture 4: Random Variables and Distributions

Normal approximation to Binomial

M378K In-Class Assignment #1

Statistics 100A Homework 5 Solutions

There are two basic kinds of random variables continuous and discrete.

Lecture 16. Lectures 1-15 Review

Evaluating Hypotheses

Basic Probability. Introduction

The Central Limit Theorem

6 The normal distribution, the central limit theorem and random samples

Stochastic Models of Manufacturing Systems

STATISTICS 1 REVISION NOTES

Week 12-13: Discrete Probability

Probability and Distributions

Discrete Random Variables. David Gerard Many slides borrowed from Linda Collins

Lecture 10: Probability distributions TUESDAY, FEBRUARY 19, 2019

Discrete Distributions

Mark Scheme (Results) Summer 2009

STA 111: Probability & Statistical Inference

ORF 245 Fundamentals of Statistics Practice Final Exam

and 1 P (w i)=1. n i N N = P (w i) lim

Bernoulli and Binomial Distributions. Notes. Bernoulli Trials. Bernoulli/Binomial Random Variables Bernoulli and Binomial Distributions.

The random variable 1

Math 105 Course Outline

Queueing Theory and Simulation. Introduction

Notes on Continuous Random Variables

Transcription:

. Introduction The quick summary, going forwards: Start with random variable X. 2 Compute the mean EX and variance 2 = varx. 3 Approximate X by the normal distribution N with mean µ = EX and standard deviation. That is, we make the approximation Pa < X < b Pa < N < b. It's not always good, but we can always make it. 4 Convert the normal distribution N to the standard normal distribution Z. Specically Z = N µ and so a µ Pa < N < b = P < Z < b µ 5 Look up the appropriate values in a table and you're done. Sometimes you just follow these steps going backwards a bit. 2. Getting an intuition for the Central Limit Theorem The following pictures are meant to illustrate the idea of the central limit theorem. Let's put say for the sake of discussion, that we are ipping an unfair coin which has the probability p = of coming up heads. If we perform just one trial, then we expect to get no heads 0 nine-tenths of the time and a heads one-tenth of the time. Figure 2. gives a histogram representation of the distribution. Notice that the height of the rectangle centered at 0 is Figure 2.. Histogram of binomial distribution with trials and probabilty p of success blue. Normal distribution with mean p and variance p p red. nine-tenths and the height of the rectangle centered at is one-tenth. If we were to use a normal distribution to approximate the rst box, we'd nd the area under the normal distribution from 0.5 to 0.5. This is illustrated as computing the green shaded area in Figure 2.2.

2 Figure 2.2. The green shaded area represents the approximation of the probability of getting no heads. Now what happens when we increase the number of trials? As we ip n coins, we have the following information: Pk heads show up = n p k p n k, k for k ranging between 0 and n. The mean number of heads to show up is ES n = np and the variance is vars n = np p. For example, when we ip 50 coins, we'd expect heads to come up, on average, 50 0 = 5 times. Let's look at the corresponding mass density functions of the binomial distribution for n = 5 Figure 2.3, n = 50 Figure 2.4, and n = 00 Figure 2.5. As before, we also plot the probability density function for the normal distribution of corresponding mean and variance. For n = 5, notice that there is a probability of about 0.6 of getting no heads. We could approximate this by nding the area under the corresponding normal distribution from x = 0.5 to 0.5. Then the probability of getting more heads decreases. But then for n = 50, there is less than a probability of 0.0 for getting no heads and the probability of getting more heads increases at rst, peaking at the probability to get 5 heads, and then decreasing. In general, as n increases, the normal distribution becomes a better and better approximation for S n. Note also that the approximation of the normal distribution to the binomial distribution for n = above is relatively bad compared to the case for n = 00. Let's continue to the next section with an example.

3 Figure 2.3. Mass density function of the binomial distribution with 5 trials and probabilty p of success blue truncated at n = 3. Probability density function of the normal distribution with mean 5p and variance 5p p red. Figure 2.4. Mass density function of the binomial distribution with 50 trials and probabilty p of success blue truncated at n = 3. Probability density function of the normal distribution with mean 50p and variance 50p p red. Figure 2.5. Mass density function of the the binomial distribution with 00 trials and probabilty p of success blue truncated at n = 22. Probability density function of the normal distribution with mean 00p and variance 00p p red.

4 Figure 3.. Histogram of binomial distribution with 200 trials and probabilty p of success blue. Normal distribution with mean 200p and variance 200p p red. Shaded in blue is the probability of getting at least 23 heads in 200 ips of a coin with probability of success p. Figure 3.2. Histogram of binomial distribution with 200 trials and probabilty p of success blue. Normal distribution with mean 200p and variance 200p p red. Shaded in red is the approximation of getting at least 23 heads in 200 ips of a coin with probability of success p. 3. Doing an Example Let's nd the probability of getting at least 23 heads after ipping a coin 200 times provided the coin has a probability of success coming up heads p =. Thus we're looking for the 0 shaded blue area in Figure 3.. But instead of using the formula for binomial distribution and adding up many terms, we'll use the normal curve to approximate. Thus, we'll be looking for the shaded red area in Figure 3.2. Note that the red region starts at 22.5 to improve the approximation. But what are we to do, because all we have is a table for the standard normal distribution! The key is in computing z-values! z = x µ,

where µ is the mean and is the standard deviation. You can choose to read some theory in Section 3. or head straight to Section 3.2 to continue with the example. 3.. Theory Behind z-value. We start o with a normal distribution X with mean µ and standard deviation, and want to nd the probability that X lies between a and b. For example, we have Figure 3.3a. By denition we have Pa X b = = ˆ b a ˆ b a fxdx 2π e x µ2 /22 dx. First let's make a u-substitution, u = x µ. Then du = dx and we have ˆ b a 2π e x µ2 /22 dx = ˆ b µ a µ 2π e u2 /22 du. Next, let's make another substitition, v = u. Thus dv = du and we have ˆ b µ a µ ˆ b µ/ /22 du = 2π e u2 a µ/ a µ = P 2π e v2 /2 dv Z b µ, where Z is the standard normal distribution mean 0 and standard deviation. Summarizing, given a normal distribution X with mean µ and standard deviation, a µ Pa X b = P X µ b µ a µ = P Z b µ where Z = X µ is the standard normal distribution. We graph the new labeling in Figure 3.3b. Notice that the shape of the curve and the area we're looking to compute remains the same as Figure 3.3a. 3.2. Resuming the Example. At the beginning of the example, we established that the probability of ipping at least 23 heads after 200 coin ips with probality is approximately 0 the area under the normal distribution X with mean 200 = 20 and variance 200 9 = 8 0 0 0 so standard deviation 3 2. In short, we have By using z-values, see theory above, we have PS 200 23 PX 22.5. PX 22.5 = P Z 22.5 20 3 2 5

6 a Normal distribution X with mean µ = 0 and standard deviation = 3. Shaded in red is probability that X lies between a = 7 and b = 5 b Standard normal distribution Z mean 0 and standard deviation. Shaded in red is probability that X lies between 7 0 3 = and 5 0 3 = 5 3. Note that to fully imitate the look of Figure 3.3a, the y-axis has been shifted from intersecting the x-axis at 0 to intersecting it at 0 0 3 = 0 3. Figure 3.3

Approximating the z-value 22.5 20 3 as 0.589..., we look in the chart to nd PZ 0.59 = 2 0.7224 and we conclude 27.5 20 P Z 3 0.7224 = 0.2776. 2 If we were to replace the numbers in this example with letters, we'd have the following. PS n k PX k 0.5 where X has mean np and variance np p and so standard deviation np p. Using z-values, we have k 0.5 np PX k 0.5 = P Z np p 7 where Z = X np np p is the standard normal distribution. 3.3. Some more theory. The end of the last example leads up the next discussion, where we're interested in X = X, where X is the total number of heads that appear after n n tosses. We have and µ = EX = n EX = n np = p varx = n varx = p p np p = 2 n2 n so that p p =. n Then we approximate, for large n, a µ Pa X b P Z b µ n a p = P Z b p n, p p p p where Z = n X p p p In any case, all we're ever doing is approximating rst by a normal distribution with the mean and standard deviation of the original distribution. Then we convert the normal distribution to a standard normal distribution.

4. Example, Going Backwards Let's apply what we've learned going backwards. Instead of being given the probability, we want to determine how often we have to ip a coin to know the probability of heads coming up within 0. of its true value with probability at least 0.8. That is, for what n do we have P Xn p 0. 0.8. Well let's rewrite the left-hand side to match what we've been discussing. P Xn p 0. = P 0. Xn p 0. = P 0. + p X n 0. + p Great, now we have our random variable between two values and we're ready to approximate it by a normal distribution. We subtract its mean p and divide by its standard deviation p p n and obtain n 0. P Z 0. n 0.8. p p p p Now part of going backwards is to nd the value c such that P c Z c = 0.8. We can use our table to do this. Using a table that gives values from to z for z 0, we have 0.2 = 0.8 and dividing by two gives us 0., so that we look for the value 0.9 in the table and nd.29 rounding up, instead of down. We have Then we need P.29 Z.29 = 0.905 0.0985 = 0.8030. n 0. p p.29 8 which is the same as n 2.9 p p or n 2.9 2 p p. Since we don't know p, the worst-case scenario is when p =, so we need 2 n 2.9 2 4.6 2 2