st: Re: Nice Margins programming challenge

[Robert Duval <rduval@gmail.com>]

I think the answer can be found in here

http://www.ats.ucla.edu/stat/stata/faq/


> Date: Fri, 3 Dec 2010 17:52:20 -0000
> From: Stephen Kay <stephen.kay@adelphigroup.com>
> Subject: st: Nice Margins programming challenge
>
> This message is in MIME format. Since your mail reader does not understand
> this format, some or all of this message may not be legible.
>
> - ------_=_NextPart_001_01CB9312.D55BA0FC
> Content-Type: text/plain
>
> I'm using MFP command combined with xtgee to model a fractional logistic
> model (dependent variable is a percent varying from 0 to 100% inclusive;
> patients are clustered within a doctor). Command is:
>
>
>
>                 mfp,select(0.05): xtgee actimpair  age  bmi comorbidities
> female
>
>
>
> The mfp command creates new tranformed age variables (lets call them age1
> and age2) based on the fractional polynomial terms it thinks best (and very
> kindly gives the formula so they are easy to reproduce).
>
>
>
> I want to run Stata 11's Margins command to give predictions on the
> dependent variable with 95% CI's across a large range of age values.
>
>
>
> It looks (to me at least) that I need to program a routine that for every
> age value needing a prediction:
>
> (1)   Generate the appropriate age1 and age2  values
>
> (2)   Input these into the margins command using the "at" option
>
> (3)   Store/append the resultant prediction plus SE, plus relevant age value
> into a matrix holding all such results.
>
> (4)   Compute the 95% CI's for the predictions and store them in the
> relevant place in the matrix mentioned in (3).
>
>
>
> Final tasks after looping through all such ages populating the matrix
> holding relevant results is to convert the matrix into data and graph from
> it (this task at least I can do).
>
>
>
> I don't program often enough to accomplish the steps above quickly.
> Hopefully some bright, public spirited spark can do it for me or show how it
> can be done even more efficiently. To be honest I'm surprised I've not found
> any previous posts or web files on this as I'd have thought quite a lot of
> people would want to do this sort of thing.
>
>
>
> Any help much appreciated.
>
>
>
> Thanks,
>
>
>
> Steve
>

*
*   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/