2015/10/19
Tools
(methods) for econometric modeling in Int. Finance
1.
Classical assumption on residuals of OLS
(Gauss – Markov theorem)
à to ensure the randomness of OLS residuals
à in multiple regression: X and Z should be independent
à to ensure the randomness of OLS residuals
à in multiple regression: X and Z should be independent
2.
Testing classical assumptions on residuals
à E(ut) = 0 (ignored if there is an intercept in OLS)
à E(ut) = 0 (ignored if there is an intercept in OLS)
è
Cov(ut,ut-j)
= 0 for j¹1 (no autocorrelation in ut)
è
Var(ut)
= su for all t (Homoskedasticity)
è
Jarque-Bera
(JB) test (Test if the residual
is normally distributed)
3.
Doornick-Hansen
test (for multivariate normality)
4.
Statistical significance testing P-values
- Choose the certain a (level of significance), for example 5%
- If p-value > a, fail to reject Ho
- If p-value < a, reject Ho
- Choose the certain a (level of significance), for example 5%
- If p-value > a, fail to reject Ho
- If p-value < a, reject Ho
5.
Testing for no-autocorrelation (Q Tests)
à H0: there is no autocorrelation among residuals (ut) up to lag p
- If reject H0, there exists autocorrelation among ut and ut-j
à H0: there is no autocorrelation among residuals (ut) up to lag p
- If reject H0, there exists autocorrelation among ut and ut-j
- If fail to reject H0, no autocorrelation among ut and ut-j
6.
Testing for homeskedasticity (Q2 tests)
à H0: there is no ARCH-type heteroscedasticity among residuals (ut) up to lag p
- If reject H0, there exists (ARCH-type) heteroskedasticity among ut
- If fail to reject H0, homeskedasticity, no (ARCH-type) heteroskedasticity among ut
à H0: there is no ARCH-type heteroscedasticity among residuals (ut) up to lag p
- If reject H0, there exists (ARCH-type) heteroskedasticity among ut
- If fail to reject H0, homeskedasticity, no (ARCH-type) heteroskedasticity among ut
7.
JB Normality test on residuals
ex: Jarque-bera test = 0.310095, with P-Value 0.856374 we can say that P-Value of JB test suggest failure to reject HO. It means that “residuals are normally distributed”
ex: Jarque-bera test = 0.310095, with P-Value 0.856374 we can say that P-Value of JB test suggest failure to reject HO. It means that “residuals are normally distributed”
Notes:
* Check OLS residuals before estimating the result of OLS in order to get BLUE estimates
BLUE means: Best, Liniear, Unbiased, Estimators.
* make sure that there are no autocorrelation, homoscesadticity, in small sample (N<30)
* Implication of the autocorrelation in residuals à Estimated coefficients are not BLUE
* Possible rescues for problems in OLS diagnosis
- Re-estimate OLS with “robut standard errors”
- model/Time sereies/AR(1)
- Re-specify the model
* Check OLS residuals before estimating the result of OLS in order to get BLUE estimates
BLUE means: Best, Liniear, Unbiased, Estimators.
* make sure that there are no autocorrelation, homoscesadticity, in small sample (N<30)
* Implication of the autocorrelation in residuals à Estimated coefficients are not BLUE
* Possible rescues for problems in OLS diagnosis
- Re-estimate OLS with “robut standard errors”
- model/Time sereies/AR(1)
- Re-specify the model
News Related to International Finance:
1.
Discordant Financial Messages From China Spur
Global Unease
(http://www.nytimes.com/2015/10/20/business/international/discordant-financial-messages-from-china-spur-global-unease.html?_r=0)
(http://www.nytimes.com/2015/10/20/business/international/discordant-financial-messages-from-china-spur-global-unease.html?_r=0)
2.
Federal Reserve Beige Book Reveals Strong US
Dollar Is Restraining Manufacturing Activity
沒有留言:
張貼留言