[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: correlations panel data
Nick, thanks for your answer. Yes, I agree that an assumption of independence
may perhaps not seem very plausible, but I don't want to impose any specific kind of
dependence structure. So, assuming that independece, (which I guess is what every
people suppose in papers when they show the correlation matrix among independent
variables in a panel and talk about the correlations being siginificant or not), how could I
test the significance?
I know that for time series you can use pwcorr, which also gives you the
significance with option sig, but it can't be used with panel data. Which could be the alternative to pwcorr, sig, but for panel data?
Thank you again very much for your help.
>As you have panel i.e. time series data,
>getting a handle on the significance
>of correlations between variables would
>seem to require either an implausible
>assumption of independence or some
>sort of specification of
>dependence structure. Which are you
>thinking of, or I am missing your point?
> I need to calculate the correlation matrix among the X
> variables in my panel dataset and also if these correlations
> are significant or not. I know there a lot of possibilities
> in time series or cross-section, but not for panel data
> (unless mvcorr, but I don't want a moving window). I could do
> the following:
> xtreg x1 x2, re
> local bx = _b[x2]
> xtreg x2 x1, re
> local by = _b[x1]
> local r = (`bx'*`by')^0.5
> macro list _r
> However, I have 20 variables, so proceeding this way could be
> very long What is more, this doesn't give me information on
> the significance.
> What could I do?
> Thanks a lot for your help in advance.
íSÚ listo! Contrata el antivirus http://www.pandasoftware.es/tienda/?idpers=109&track=13920