Financial professional with expertise in financial modeling, risk management, and related quantitative methods including Monte Carlo simulation, linear and logit regression, and statistical hypotheses testing. For past few years have been engaged in developing econometrics models related to Stress testing for a large financial institution.

May
27
awarded Student
May
26
awarded Popular Question
May
15
awarded Notable Question
Mar
31
answered Should I use the t-test or the Wilcoxon rank sum test given these qqplots and Shapiro-Wilk stats
Mar
31
comment In R Studio, how to save your graphs?
rawr, so you suggest that save.image is better than going through the Project interface?
Mar
30
comment In R Studio, how to save your graphs?
r2evans, those are good leads. Thanks
Mar
30
asked In R Studio, how to save your graphs?
Mar
30
accepted When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Mar
30
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Richard, I saw your answer and comment. And, I am not sure if I understand them. That's because if the variance is unobservable and is replaced by the proxy Residual^2... then on the righthand side of the equation, both the ARCH term = GARCH terms which does not sound right.
Mar
29
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Richard, I'll read the article. But, if you find it difficult, I most probably will find it obtuse. So, how does GARCH structures the dependent variable? I thought I read somewhere that it does simply take Residual^2 as a proxy for the residual's variance. But, does that sound right to you?
Mar
28
awarded Informed
Mar
28
awarded Autobiographer
Mar
27
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Actually, this is a piece of the GARCH(1,1) model I really don't understand. What is the dependent variable? It is the variance of Y. But, how do you calculate it for every single observations within the learning sample?
Mar
27
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Richard, I give you +1 not only for a very good answer but also some excellent comments. Anyone, before contributing on this issue is encouraged to read Richard's comments first. I am not giving points for an answer that Richard has already fully addressed in the comment section.
Mar
27
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Richard, you understand my model framework precisely. Just a couple of details. All the original regressors have positive reg. coefficients within the original regression or Mean Equation of GARCH. Out of four exogenous regressors (in my original model, not the example) only one is stat. significant. The ARCH term is not sign. and has flipped signs within the GARCH(1,1) + exogenous variables. In the GARCH(1,1) the ARCH term is positive but not stat. sign. The GARCH term is not stat sign. in GARCH (1,1). But, it is in GARCH(1,1) + ex. variables.
Mar
27
comment Is it ok to have a unit root within an independent variable?
Cagdas and Richard, excellent comments. Thanks much. I have up-flagged them all. Cagdas, I have no experience with Cointegration models; but, intuitively I so agree with you. They seem to me like unicorns. I can't see how two "level" variables could possibly have a difference between them that is stationary for long enough to build a model with. Your State Space model is intriguing. I'll have to study that. Richard my econometric model example was just that, facilitate our conversation. I am not suggesting there is any theory behind it suggesting they could be cointegrated.
Mar
27
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Richard, I am working on a company proprietary model so I can't share the specifics. I'll use an example that replicates the situation. GNP ~ S&P500 + Housing price. So, when I run a GARCH model (using EViews)... I first run it using only the ARCH and GARCH terms. And, neither is statistically significant within the Variance Equation. I run the GARCH model a second time, with the two independent variables. Now, the GARCH term is very stat. significant and so is of the ind. variable. What do you conclude in this situation?
Mar
27
comment Is it ok to have a unit root within an independent variable?
Richard, we can make any hypothetical examples which would work just fine. Let's say your have a model: GNP ~ S&P500 + Housing price + 10 Yr. Treasury rates. If you take all those variables on a nominal level basis, they all have huge Unit roots. Three of them trend upward forever (GNP, S&P500, Housing price). You could detrend them all by looking at periodic % change (from one quarter to the next) and avoid the Unit root issue. But, maybe you could keep them on a nominal basis (Cointegration framework). I don't know.
Mar
27
comment When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
Your comment makes perfect sense to me. Is the heteroskedasticity associated with the level of the fitted value or estimated value? Or do you want to go more granular and check whether the heteroskedasticity is associated with a particular independent variable(s). In my mind, your comment qualifies as a good answer. I would give it thumbs up. You are welcome to structure it as such.
Mar
27
asked When testing autoregressive conditional heteroskedasticity with GARCH do you need to include the ind. variables?
1 2 3 4 5