Your question seems analogous to
. sysuse auto, clear
. tab for rep78 , taub
| Repair Record 1978
Car type | 1 2 3 4 5 | Total
-----------+-------------------------------------------------------+----------
Domestic | 200 800 2,700 900 200 | 4,800
Foreign | 0 0 300 900 900 | 2,100
-----------+-------------------------------------------------------+----------
Total | 200 800 3,000 1,800 1,100 | 6,900
Kendall's tau-b = 0.5589 ASE = 0.007
. ret li
scalars:
r(N) = 6900
r(r) = 2
r(c) = 5
r(taub) = .558892945435974
r(ase_taub) = .0070958155056459
. di r(taub) / r(ase_taub)
78.763737
I guess that ratio is the z-value you want, although
some might label it -t-. Similarly, you can get a p-value
through whatever distribution and df is appropriate. You
just need to rummage around in the statistical functions.
I guess this is available more directly as one of
several special cases of Roger Newson's -somersd-,
but I could be wrong.
Nick
[email protected]
Deidra Young
> I am interested in a simple solution to asking Stata to calculate the
> p-value and z statistic for comparing two groups by ordinal
> categorical
> variables (mostly three to five categories). When I use the
> the following
> (in Stata 9):
>
> tab autism kerr, all
>
> Where autism is a variable denoting two groups of people and kerr is a
> ordered severity score for a clinical symptom, the chi square test is
> inappropriate due to the ordering of kerr (0 1 2 3) ranking
> the severity of
> symptoms. This means that the Kendall tau-b score is more
> appropriate as a
> measure and the approximate standard error is provided. But how can I
> insert a command to request the calculation of the p value
> for this (similar
> to the ktau procedure, but in the tab context above).
>
> I have been working on this today and am perplexed about how
> to automate
> this. There seems to be an ado file for this available, but that is a
> little more complicated than necessary. Also, I am new to
> ado files (not
> tried one yet but have been looking at how to do these).
>
> I wondered if I should try the new tabout.ado but I must print off the
> tutorial first. It looks like a good program! (but not for
> producing p
> values).
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
