End-Semester Examination MA 373 : Statistical Analysis on Financial Data

Transcription

1 End-Semester Examination MA 373 : Statistical Analysis on Financial Data Instructor: Dr. Arabin Kumar Dey, Department of Mathematics, IIT Guwahati Note: Use the results in Section- III: Data Analysis using R to answer the questions in Part- I. Writing the name of any theorem is not sufficient. State the theorem clearly before using it. ALL tests use the 5% significance level, and all VaR calculations use 1% tail probability. Round your answer to 3 significant digits. Total 80 marks. 1 Part - I Consider the daily log returns of the Starbucks stock, in percentages, from January 1993 to December The relevant R output is attached. Answer the following questions. 1. (2 points) Is the mean log return significantly different from zero? Why? mark 1

2 2. (2 points) Is there any serial correlation in the log return series? Why? 3. (2 points) An MA model is used to handle the mean equation, which appears to be adequate. Is there any ARCH effect in the return series? Why? 4. (4 points) A GARCH(1,1) model with Student-t distribution is used for the volatility equation. Write down the fitted model, including the degrees of freedom of the Studentt innovations and mean equation. 5. (4 points) Since the constant term of the GARCH(1,1) model is not significantly different from zero at the 1% level, an IGARCH(1,1) model is used. Write down the fitted IGARCH(1,1) model, including the mean equation. 6. (3 points) Is the IGARCH(1,1) model adequate? Why? What is the 3-step ahead volatility forecast with the last data point as the forecast origin? 2 Part - II 1. (5 points) Describe non-parametric procedure for estimating shape parameter of GEV distribution. Describe two drawbacks of this procedure. 2. (2 points) What will be the VaR of a portfolio consists of a stock and a option of that stock? 3. (5 points) Suppose, X 1, X 2, X 3 are coming from the following AR(1) process: X t = φ 1 X t 1 + ɛ t with φ 1 < 1. Also assume X 2 is missing and you have information of X 1 and X 3. Predict X 2 and find out the expression of corresponding mean square error. 4. (4 points) Show that there is no stationary solution of the difference equations X t = φx t 1 + Z t, Z t W N(0, σ 2 ) if φ = ±1. 2

3 5. (8 points) Show that in order for an AR(2) process with autoregressive polynomial φ(z) = 1 φ 1 z φ 2 z 2 to be causal, the parameters (φ 1, φ 2 ) must lie in the triangular region determined by the intersection of the three regions, φ 2 + φ 1 < 1, φ 2 φ 1 < 1 and φ 2 < (8 points) Find the mean and auto-covariance function of the ARMA(2,1) process, X t = X t 1 0.4X t 2 + Z t + Z t 1, {Z t } W N(0, σ 2 ). Is the process causal and invertible? 7. (8 points) For MA(2) process find the largest possible values of ρ(1) and ρ(2). 8. (7 points) Find out the excess kurtosis for conditional Gaussian GARCH(1,1) model and make proper interpretation for the property. 9. (a) (6 points) Show that if Σ is the covariance matrix of the random vector X = (X 1, X 2,, X p ) and the eigenvalue-eigenvector pairs of Σ are (λ 1, e 1 ), (λ 2, e 2 ),, (λ p, e p ) such that λ 1 λ 2 λ p, the i-th Principal Component of X is given by Y i = e ix = e i1 X 1 + e i2 X e ip X p i = 1(1)p. Furthermore, V (Y i ) = e iσe i = λ i i = 1(1)p. Cov(Y i, Y j ) = e iσe j = 0, i j (b) (3 points) State three objectives of principle component analysis. 10. (3 points) What will be copula form of the following joint density function : { 1 e x e y + e x y θxy if x 0, y 0 H θ (x, y) = 0, otherwise (4 points) How do you generate random number from H θ (x, y) using the above copula? 3

4 3 Data Analysis using R : ***** PROBLEM - PART-I************************ > x =read.table( d-sbuxsp9307.txt ) > sbux=log(x[, 2] + 1) 100 > basicstats(sbux) nobs NAs Minimum Maximum Mean SE Mean LCL Mean UCL Mean Variance Stdev Skewness Kurtosis > Box.test(sbux,lag=15,type= Ljung ) Box-Ljung test data: sbux X-squared = , df = 15, p-value = > acf(sbux) Not shown, but shows two significant ACFs at lags 1 and 2. > m1=arima(sbux,order=c(0,0,2)) > m1 Call: arima(x = sbux, order = c(0, 0, 2)) 4

5 Table 1: Coefficients: ma1 ma2 intercept s.e σ 2 estimated as 7.275: log likelihood = , aic = > Box.test(m1$residuals,lag=15,type= Ljung ) Box-Ljung test data: m1residuals X-squared = , df = 15, p-value = > Box.test(m1$residuals 2,lag=15,type= Ljung ) Box-Ljung test data: m1$ residuals 2 X-squared = , df = 15, p-value < 2.2e-16 5

6 Table 2: Maximum Likelihood Estimation (Standard Errors based on Second derivatives) Coefficient Std.Error t-value t-prob Cst(M) MA(1) MA(2) Cst(V) ARCH(Alpha1) GARCH(Beta1) Student(DF) No. Observations : 3778 Parameters : 7 Mean (Y) : Variance (Y) : Skewness (Y) : Kurtosis (Y) : Log Likelihood : Alpha[1]+Beta[1]: Table 3: Other results > v1=garchoxfit(formula.mean= arma(0,2),formula.var= garch(1,1),series=sbux, cond.dist= t ) ******************** **SPECIFICATIONS ** ******************** Mean Equation : ARMA (0, 2) model. No regressor in the mean Variance Equation : GARCH (1, 1) model. No regressor in the variance The distribution is a Student distribution, with degrees of freedom. See Table 2 for Maximum Likelihood Estimation (Standard Errors based on Second derivatives). See Table 3 for Other results. 6

7 ** TESTS ************* Information Criterium (to be minimized) Akaike Shibata Schwarz Hannan-Quinn Q-Statistics on Standardized Residuals P-values adjusted by 2 degree(s) of freedom Q( 10) = [ ] Q( 15) = [ ] Q( 20) = [ ] H 0 : No serial correlation Accept H 0 when prob. is High [Q < Chisq(lag)] Q-Statistics on Squared Standardized Residuals P-values adjusted by 2 degree(s) of freedom Q(10) = [ ] Q(15) = [ ] Q(20) = [ ] H 0 : No serial correlation Accept H 0 when prob. is High [Q < Chisq(lag)] 7

8 ARCH 1-2 test: F(2,3771)= [0.8556] ARCH 1-5 test: F(5,3765)= [0.6782] ARCH 1-10 test: F(10,3755)= [0.8247] 8

9 Table 4: Maximum Likelihood Estimation (Std.Errors based on Second derivatives) Coefficient Std.Error t-value t-prob Cst(M) MA(1) MA(2) ARCH(Alpha1) GARCH(Beta1) > v2=garchoxfit(formula.mean= arma(0,2),formula.var= igarch(1,1), series=sbux,include.var=f) ******************** ** SPECIFICATIONS ** ******************** Mean Equation : ARMA (0, 2) model. No regressor in the mean Variance Equation : IGARCH (1, 1) model. No regressor in the variance. The distribution is a Gauss distribution. See Table 4 for Maximum Likelihood Estimation (Std.Errors based on Second derivatives). Log Likelihood :

10 Horizon Mean Variance ????? ????? Table 5: FORECASTS ******************* ** FORECASTS ** ******************* Number of Forecasts: 15 See Table 5 for different forecast values. *********** ** TESTS ** *********** Information Criterium (to be minimized) Akaike Shibata Schwarz Hannan-Quinn Q-Statistics on Standardized Residuals P-values adjusted by 2 degree(s) of freedom Q( 10) = [ ] Q( 15) = [ ] Q( 20) = [ ] H 0 : No serial correlation Accept H 0 when prob. is High [Q < Chisq(lag)] 10

11 ARCH 1-2 test: F(2,3771)= [0.7744] ARCH 1-5 test: F(5,3765)= [0.3909] ARCH 1-10 test: F(10,3755)= [0.6775] Q-Statistics on Squared Standardized Residuals P-values adjusted by 2 degree(s) of freedom Q( 10) = [ ] Q( 15) = [ ] Q( 20) = [ ] H 0 : No serial correlation Accept H 0 when prob. is High [Q < Chisq(lag)] 11