Does Pcorr not store significance values of partial correlations?

[Maithreyi Gopalan]

Hi there,

It's been a while since I posted a question here - apologize if I've forgotten (missed) some of the norms. I am writing to ask if the function "pcorr" really does not store the significance values of the partial correlations in matrices? The help file lists just the following as being part of the stored results.

pcorr stores the following in r():

Scalars
r(N) number of observations
r(df) degrees of freedom

Matrices
r(p_corr) partial correlation coefficient vector
r(sp_corr) semipartial correlation coefficient vector

Am I missing something obvious? Thanks for your help!

Best,
Maithreyi

[daniel klein]

Isn't the p-value simply the p-value from the underlying regression model?

To elaborate using an example:

. sysuse auto
(1978 automobile data)

. pcorr price mpg turn gear_ratio
(obs=74)

Partial and semipartial correlations of price with

               Partial   Semipartial      Partial   Semipartial   Significance
   Variable |    corr.         corr.      corr.^2       corr.^2          value
------------+-----------------------------------------------------------------
        mpg |  -0.3497       -0.3288       0.1223        0.1081         0.0026
       turn |  -0.0662       -0.0585       0.0044        0.0034         0.5804
 gear_ratio |  -0.0607       -0.0535       0.0037        0.0029         0.6127

. regress price mpg turn gear_ratio , noheader pformat(%5.4f)
------------------------------------------------------------------------------
       price | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
         mpg |  -249.3979   79.86385    -3.12   0.0026    -408.6813   -90.11447
        turn |  -62.36816   112.2961    -0.56   0.5804    -286.3358    161.5994
  gear_ratio |  -485.7191   955.1604    -0.51   0.6127    -2390.727    1419.288
       _cons |   15413.95   6887.792     2.24   0.0284      1676.68    29151.22
------------------------------------------------------------------------------

The p-values are stored in r(table) after regress.They are also easily re-calculated (see Buis, 2007).


Buis, M. 2007. Stata tip 53: Where did my p-values go? The Stata Journal, 7(4), 584–586

[Maithreyi Gopalan]

Thank you, daniel! In this case, I just needed the correlation coefficient and p-values for a large number of variables. So, I was hoping to directly pull those from the stored matrices. I suppose I could just run the underlying regression and back-calculate the correlation coefficient from the betas and p-values from r(table). I just found it odd though as it would have been wonderful to have r(sig) stored in the case of pcorr too just like with pwcorr! Thank you again!

[daniel klein]

You could probably modify pcorr.ado accoringly. Or, you can set up a wrapper:

program pcorr_sig , rclass
    
    version 17
    
    syntax varlist
    
    quietly regress `varlist'
    
    tempname p
    matrix `p' = r(table)[4, 1..e(df_m)]'
    
    pcorr `varlist'
    
    return matrix sig = `p'
    
    return add
    
end

Here is an example:

. pcorr_sig price turn mpg gear_ratio
(obs=74)

Partial and semipartial correlations of price with

               Partial   Semipartial      Partial   Semipartial   Significance
   Variable |    corr.         corr.      corr.^2       corr.^2          value
------------+-----------------------------------------------------------------
       turn |  -0.0662       -0.0585       0.0044        0.0034         0.5804
        mpg |  -0.3497       -0.3288       0.1223        0.1081         0.0026
 gear_ratio |  -0.0607       -0.0535       0.0037        0.0029         0.6127

.
. return list

scalars:
                 r(df) =  70
                  r(N) =  74

matrices:
            r(sp_corr) :  3 x 1
             r(p_corr) :  3 x 1
                r(sig) :  3 x 1

.
. matlist r(sig)

             |    pvalue
-------------+-----------
        turn |  .5803986
         mpg |  .0026049
  gear_ratio |   .612687

Note that the wrapper is just a proof of concept. It does not handle factor variables, omitted variables (due to collinearity), etc. Do not use it blindly.

[Maithreyi Gopalan]

Yayy - thank you so much daniel! This is great.