How to make an (inverted) covariance matrix a full matrix

[Todd Jones]

I would like to create a full covariance matrix and then invert the matrix.

Say that I compute the covariance as follows:

sysuse auto2, clear  
corr price mpg headroom trunk weight length, covariance

This produces:

             |    price      mpg headroom    trunk   weight   length
-------------+------------------------------------------------------
       price |  8.7e+06
         mpg | -7996.28   33.472
    headroom |  285.721 -2.02536  .715707
       trunk |  3965.67 -14.3924  2.39559  18.2962
      weight |  1.2e+06 -3629.43  317.873  2234.66   604030
      length |  28360.3 -102.514  9.72556  69.2025  16370.9   495.79

My main problem is how to instead get it to compute the matrix such that each cell is filled in. In other words, I would like to create a full covariance matrix instead of the lower triangular matrix.

And then would this be the correct way to invert the matrix?

matrix a = r(C)
matrix b = invsym(a)
matrix list b

This produces:

symmetric b[6,6]
               price         mpg    headroom       trunk      weight      length
   price   1.858e-07
     mpg   .00001593   .08966729
headroom   .00013193   .00988956   2.5935193
   trunk  -.00002065   .00125248  -.31581775   .15498871
  weight  -8.212e-07   .00027199  -.00069926   .00020868   .00002079
  length   .00002008   .00827906   .01079437  -.02088888  -.00059878   .02505605

[Mike Lacy]

Your symmetric matrices a and b do contain entries in each cell. Stata doesn't display the upper triangle for symmetric matrices. See -help matrix list- with attention to the -nohalf- option.

[Todd Jones]

Ah, I had no idea. Thank you very much!