EViews.com

Hi there, I have a quick question regarding the heteroskedasticity tests in eviews 6 student version. My data contains of 365 datapoints, 27 endogenous variables (some of them are differenced and/or logged and some are normal), 13 exogenous variables and 2 lags (the number of lags are not ideal but more lags are not possible due to the nr of variables versus the number of datapoints). When trying to conduct the white test with cross-terms I get the following error message: positive or non-negative argument to function expected. I searched the whole internet but I simply can not find an answer why eviews is giving me this error.

As a solution I computed the white test myself, as for the white test the obs*rsquared is used. So I multiplied all the rsqaures of the 27 variables with the observations (which were 356) and when that value is higher than the chi square the data is heteroskedastic. Something that we don't want to happen. And now I came to the biggest suprise, as the chi square is 410 (356 observations, zp=1.96) and the obs*squared can never ever be larger than 356, and that only happens when the r square is 1... So now it looks like all of my variables are perfectly homoskedastic but I simply don"t believe it. Especially when looking at the scatterplots of the residuals.

Hopefully some of you can find a mistake I have made or tell me that indeed all of my variables are homoskedastic.

Thanks in advance

You'll have to provide more detail on your specifications, or even better yet, the workfile with an example of an equation you are trying to test...

As to the auxiliary regression results, again it's hard to say without seeing exactly what you are doing...

Thanks for your quick reply. Unfortunately I am not allowed to post the workfile online.