egen corr() with weights

[John Hanser]

Dear Statalisters,

I have panel data and want to compute a time series of cross-sectional correlations of two variables, x1 and x2.

So far, my approach has been:

bysort year: egen correlation = corr(x1 x2)

However, the corr() command from egenmore does not allow weights. I am struggling to find an efficient work around...

Any suggestions are highly appreciated!

Best,
John

[Leonardo Guizzetti]

You doing mention the type of weights but I would look at -corr- and use those returned results.

[John Hanser]

You doing mention the type of weights but I would look at -corr- and use those returned results.

Thanks for the response, Leonardo. I did exactly that now... I was just hoping there is an easier way.

[Joro Kolev]

Try this:

. sysuse auto, clear
(1978 automobile data)

. levelsof rep, local(repslevs)
1 2 3 4 5

. gen corrmpgheadroom = .
(74 missing values generated)

. foreach l of local repslevs {
  2. correlate mpg headroom [w=weight] if rep == `l'
  3. replace corrmpgheadroom = r(rho) if rep == `l'
  4. }