Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Pairwise correlation of non-continuous variables


From   Glenn Goldsmith <glenn.goldsmith@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Pairwise correlation of non-continuous variables
Date   Thu, 10 Sep 2009 15:01:46 +0100

<>

I'm not an expert on this, but seeing as no-one else has replied,
here's my 2 cents:

1. Kendall's tau seems a reasonable test for this situation. However,
you probably want to use the -stats(taub)- option rather than the
default -stats(taua)- if you have tied values, as this will deal with
them better.

2. The -p(0.05)- option is short for -print(0.05)- and only displays
correlations that are significant at the specified level. If you want
to display all the coefficients, you could instead use -star(0.05)-,
to indicate which are significant.

3. Given that you're testing multiple correlations, you may also want
to think about correcting the significance levels for this to control
the overall error rate using the -bonferroni- or -sidak- options.

HTH,

Glenn

"Sockolow, Paulina" <psockolo@jhsph.edu> wrote:

Hi Stata list,
I have data from a small survey , and wanted to test for pairwise
correlation among the items. There are less than 40 subjects, and 21
ordinal variables. I wanted to use a non-parametric test, setting
statistical significance at 0.05, and ran:
 ktau <variable list> , pw p(0.05)

1. Was this the correct test for my question?

2. The output does not look like pairwise correlation output in that
only some pairs have a value reported. Because of the ktau command,
and the values reported are between 0 and 1, I am guessing the values
are Kendall's tau. How do I interpret the output? Did Stata return
only pairs that are correlated?

Thank you,
Paulina
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index