st: RE: How to display multiple stars in correlation matrix

[Nick Cox <n.j.cox@durham.ac.uk>]

Somebody may want to help you to do this. 

My aim is different: to argue that this practice, although common, is an elaborate way of diverting attention from better analyses. 

I focus throughout on Pearson correlation, although similar issues arise for other similar measures. (For example, Roger Newson has long explained and supported better practices for various rank correlation measures and their kin.) 

First, although there are some exceptions when a correlation is really interesting in its own right (e.g. within genetics) most of the time a correlation is at best an indication, a pointer that there is a relationship that should bear closer examination, usually by looking at the corresponding regression more closely. A correlation is rarely going to be the conclusion to an interesting statistical argument, and decorating a table of correlations is to that extent a waste of effort. 

Second, if a correlation has a merit, it's as a measure of the strength, meaning in this case linearity, of a relationship and if there is uncertainty about its size, that's better tackled by calculating a confidence interval rather than testing a null hypothesis. Dedicated machinery is available through 

SJ-10-4 pr0041_1  . . . . . . . . . . . . . . . . . Software update for corrci
        (help corrci, corrcii if installed) . . . . . . . . . . . .  N. J. Cox
        Q4/10   SJ 10(4):691
        update to fix corrci so that it saves r-class results

SJ-8-3  pr0041  .  Speaking Stata: Corr. with confidence, Fisher's z revisited
        (help corrci, corrcii if installed) . . . . . . . . . . . .  N. J. Cox
        Q3/08   SJ 8(3):413--439
        reviews Fisher's z transformation and its inverse, the
        hyperbolic tangent, and reviews their use in inference
        with correlations

Third, what's the argument here? That readers need help in identifying the correlations in a matrix that are to be taken seriously, because the table is too large, or too dull, or too complicated, to be worth perusing. If that's so, why not use the space for a scatter plot matrix, as -graph matrix- will provide? 

Further fulminations against starring can be found at 

http://www.stata.com/statalist/archive/2007-03/msg00646.html

Nick 
n.j.cox@durham.ac.uk 

Alvaro Filipe da Costa

I am able to generate a Pearson correlation matrix and indicate a good
significance level with a star(*) using the pwcorr and pwcorrs commands.

Could anyone help me out with a way to display one or several stars
depending on the significance level?
E.g.: * for p<0.05 and  ** for p<0.001.


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/