标题: Correcting for ARCH [打印本页] 作者: anothercfainnyc 时间: 2011-7-11 19:28 标题: Correcting for ARCH
So, I know how to correct for traditional conditional heteroskedasticity, but how do you correct for Auto-regressive Conditional heteroskedasticity?作者: yodacaia 时间: 2011-7-11 19:28
generalized least squares, whatever that means....
from wikipedia: is a technique for estimating the unknown parameters in a linear regression model. The GLS is applied when the variances of the observations are unequal (heteroskedasticity), or when there is a certain degree of correlation between the observations. In these cases ordinary least squares can be statistically inefficient, or even give misleading inferences.作者: kingstongal 时间: 2011-7-11 19:28
Use white-corrected std errors for Conditional Heterosked
Use Hansen method to adjust std errors for Serial Correlation (auto-regressive)
Use Hansen method to adjust std errors for Cond Heterosked AND serial correl; it adjust for both.作者: kickthatcfa 时间: 2011-7-11 19:28
really you think its new?作者: ppls 时间: 2011-7-11 19:28
I should probably rephrase: I don't recall hearing about this stuff. It's possible I saw it and didn't pay attention to it, but I seriously don't remember either of those things at all. If they had shown up on the exam last year, I would've had no idea wtf they were.作者: KungFuPanda 时间: 2011-7-11 19:28
SeesFA Wrote:
-------------------------------------------------------
> Use white-corrected std errors for Conditional
> Heterosked
>
> Use Hansen method to adjust std errors for Serial
> Correlation (auto-regressive)
>
> Use Hansen method to adjust std errors for Cond
> Heterosked AND serial correl; it adjust for both.
Good post SeesFA!
Also, use stepwise regression to remove variables from regression to minimize multicollinearity.
Edited 1 time(s). Last edit at Tuesday, April 12, 2011 at 01:04AM by Iginla2010.作者: Otabek 时间: 2011-7-11 19:28
what is stepwise regression作者: President1988 时间: 2011-7-11 19:28
Correction of multicollinearity is not simple. It often results from too many variables in the model caused by "over-fitting".
In stepwise regression you examine the impact of each variable to the model step by step. The variable that cannot contribute much to the variance explained (not statistically significant) would be removed.
Edited 1 time(s). Last edit at Tuesday, April 12, 2011 at 12:44PM by Iginla2010.作者: ayodayo 时间: 2011-7-11 19:28
@magicskyfairy
good one . based on your experience in FRA what are the 'red alert' areas offcourse pension and intercorp investments but I dont understand the intenton behind the first 2 readings( Inventory valuation and long lived assets ) when we have already studied in Level 1 , that apart is that area really brutal (b/c of high weight) ? and how about equity ? I havent even started yet , even though I am a buy-side analyst I havent really used 2step or a 3 step ddm
