Re: st: permutations

[Christopher Intemann <intemann@gmail.com>]

Hi Dave
Am 13.03.2008 um 15:24 schrieb David Airey:

> .
>
> Do you have access to the manual help or online help? The syntax is:
>
> permute permvar exp_list [, options] : command
>
> You specified permvar but not exp_list, so I'm not sure what permute  
> is defaulting to, maybe just the overall model statistic? You can  
> write a small program to return the coefficients for the genetic  
> dummy variables, and an overall test of those, that gets called by  
> permute. The ANOVA example in the printed help shows how to do this.
>
> T is your statistic
> T(obs) is your observed statistic
> c is the number of times the statistics is greater than T(obs)
> n is the default number of permutations since you didn't specify  
> this either
>
> Below is an example from ATS UCLA (Stata version 9/10):
>
> * read your data into Stata
> use http://www.ats.ucla.edu/stat/stata/notes/hsb2, clear
> generate goodread = (read > 60)
>
> * run logistic predicting "goodread" from "write"
> logit goodread female
>
> * run permutation test for above with 10000 repetitions
> permute female x2=e(chi2), reps(10000) nodots: logit goodread female
>
> Since you want a statistic for the individual dummy variables and  
> the test for both genetic dummy variables, you don't just want the  
> results from the overall test, and you cannot just collect that  
> using permute as the above example does.
I was using a pre-stata9 syntax of the permute command.

However, a command like

xi:permute affected_stata x2=e(chi2):logistic affected_stata i.gene  
age ethno_x sex_x

does calculate some permutations, but is pretty sure not what I want:-(

As i could calculate my p-values from the matrices e(b) and e(V),  
which are generated by logistic, i would  rather like to use these in  
the exp_list. However, this does not work:
"matrix operators that return matrices not allowed in this context"

Is there any other way to permute or calculate the p-values?
Just similar to what other programs for genetic analysis (like  
unphased, plink or haploview) do?
I just came along cvpermute, maybe this could do the trick?
Many thanks in advance,
   Chris
*
*   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/