Re: st: Regression Diagnostics for Models with cluster-robust standard errors

[Nick Cox <njcoxstata@gmail.com>]

You have got (more) honest standard errors. They say nothing directly  
about precisely how the model might need improvement. It's now 50  
years and more since plotting residuals versus fitted was suggested as  
a basic follow-up to regression. Also, whether the functional form is  
about right is more crucial than assumptions about error terms.

Nick

On 26 Jan 2013, at 13:17, Barbara Engels <engels.ba@gmail.com> wrote:

> Dear Statalist people,
>
> I'm sorry if this is obvious, but to me, it is not:
>
> I estimated an OLS regression with cluster-robust standard errors  
> and want to perform common regression diagnostics now, i.e. check  
> for linearity, outliers, normallity, autocorrelation,  
> homoskedasticity, multicollinearity, model specification (ramsey) ....
> Does it make sense to check for homoskedasticity, normality and  
> autocorrelation at all or is it just sufficient that I accounted for  
> it with the cluster-robust s.e.?
>
> Thank you very much for your time.
>
> Best, Barbara
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/