Re: st: RE: Storing coefficients from a stepwise model

[Elaine Thomas <e.thomas@cphc.keele.ac.uk>]

Dear all

Just want to thank David Harrison for his very elegant solutions to my 
problems...David, you've saved us a lot of transcribing !
Elaine

David Harrison wrote:


> You don't say how you are "storing" the coefficients... I will assume it is with something like post, e.g.
>
> sw logit y x1 x2 x3 x4 x5, pr(.1)
> post f (_b[x1]) (_b[x2]) (_b[x3]) (_b[x4]) (_b[x5])
>
> Suppose x2 gets dropped in the stepwise procedure, then this code will produce an error as _b[x2] does not exist. Instead you need to check that each coefficient exists first, e.g.
>
> sw logit y x1 x2 x3 x4 x5, pr(.1)
> foreach v of varlist x1 x2 x3 x4 x5 {
> 	cap local b_`v'=_b[`v']
> 	if _rc {
> 		local b_`v'=.
> 	}
> }
> post f (`b_x1') (`b_x2') (`b_x3') (`b_x4') (`b_x5')
>
> Hopefully this can be adapted to your needs.
>
> David
>
> -----Original Message-----
> From: Elaine Thomas [mailto:e.thomas@cphc.keele.ac.uk]
> Sent: 01 March 2005 12:53
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Storing coefficients from a stepwise model
>
>
> Dear all
>
> I want to perform a stepwise logistic procedure on a number of 
> bootstrapped samples and store the coefficients from each application of 
> the regression in a single file.
> I have thus far been able to generate the bootstrapped samples (thanks 
> to David Harrison!) and store the coefficients - however, only if I do 
> not use a stepwise approach to the regression, but simply tell it to fit 
> a model containing a set number of independent variables, say x1 -x5, - 
> this works well. So I get a single new file which contains the 
> coefficients for each cycle, ie
>
> Sample         x1        x2        x3        x4        x5
>   1             0.24    0.35     0.84    0.08     0.75
>   2             0.22    0.40     0.78    0.12     0.69 ......
>
> The complication in using the sw procedure is that the model derived in 
> each sample may contain anywhere from all to none of the variables 
> offered and when I try to get Stata to store the coefficients for the 
> offered vbls, say x1-x5, I can't get it to store "." for those vbls that 
> are not chosen in that cycle of the procedure
>
> What I want is the same form of file as above but with . or even 0 for 
> those vbls not ending up in the model for that sample, ie.
>
> Sample         x1        x2        x3        x4        x5
>   1             0.24        0      0.84    0.08     0.75
>   2             0.22    0.40     0.78        0      0.69 ......
>
> Any help would be gratefully received
>
> Best wishes
> Elaine Thomas
>
>
> *
> *   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/
>  
--
Dr Elaine Thomas
Senior lecturer in Biostatistics
Primary Care Sciences Research Centre
Keele University
North Staffordshire
ST5 5BG

Telephone : + 44(0)1782 583924
Fax       : + 44(0)1782 583911        
Email     : e.thomas@keele.ac.uk





*
*   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/