Hello 
I'm running a panel data regression, with 5 independent variables and 28 firms over 5 years. I have 140 observations.
After running a Hausman test, i found that a FE reg is to be used. Next I tested for heteroscedasticity - using the Cook-Weisberg httest for residuals - and autocorrelation - using the xtserial command for panel data. Both turned positive. My data is characterized by both heteroscedasticity and autocorrelation.
I then looked for ways to correct for them. I learned the following:
>> heteroscedasticity - use robust (eg. xtreg dep, var1, var2....., fe vce(robust))
>> autocorrelation - use Cochranne Orcutt method (prais dep, var1, var2...., corc)
But I need to correct them simultaneously in a single regression. I did found something using google :the Newey-west method...
I ran newey dep, var1, var2....., lag (1) force
My questions and problems are as follows:
> How to obtain r-squared when running newey?
> is there another more efficient way of correcting for both autocorrelation and heteroscedasticity?
> should i run dfuller to ensure stationarity? if yes, how to interpret dfuller please?
Any help would be much appreciated....

I'm running a panel data regression, with 5 independent variables and 28 firms over 5 years. I have 140 observations.
After running a Hausman test, i found that a FE reg is to be used. Next I tested for heteroscedasticity - using the Cook-Weisberg httest for residuals - and autocorrelation - using the xtserial command for panel data. Both turned positive. My data is characterized by both heteroscedasticity and autocorrelation.
I then looked for ways to correct for them. I learned the following:
>> heteroscedasticity - use robust (eg. xtreg dep, var1, var2....., fe vce(robust))
>> autocorrelation - use Cochranne Orcutt method (prais dep, var1, var2...., corc)
But I need to correct them simultaneously in a single regression. I did found something using google :the Newey-west method...
I ran newey dep, var1, var2....., lag (1) force
My questions and problems are as follows:
> How to obtain r-squared when running newey?
> is there another more efficient way of correcting for both autocorrelation and heteroscedasticity?
> should i run dfuller to ensure stationarity? if yes, how to interpret dfuller please?
Any help would be much appreciated....
Comment