Re: st: How to make prediction after a spline regression

[Roger Newson <roger.newson@kcl.ac.uk>]

At 18:40 13/04/04 -0700, Shige Song wrote:
> Dear Roger,
>
> Thanks for the reply. One more thing, is the "noconstant" option in the 
> logistic regression mandatory or optional? I did not include in my xtlogit 
> model, it did not seem to influence the estimate (none of the spline terms 
> was dropped).
The -noconstant- option is mandatory if you want the parameters 
corresponding to the reference splines to be the odds at the reference 
point. If you do not have a -noconstant- option, then you should also omit 
the reference spline corresponding to a reference point inside the 
completeness range, and the parameters corresponding to the other reference 
spline will then be odds ratios, comparing the values of the spline at the 
other reference point to the value of the spline at the ommitted reference 
point. However, in this case, the value of the spline at the ommitted 
reference point will be given by the baseline odds parameter _cons, which 
is usually not printed by Stata.

The best thing to do is probably to use the -noconstant- option with a unit 
variable. For instance, if the reference splines are cs_1, cs_2, ... , 
cs_10, and the ommitted reference spline is cs_5, and the binary outcome 
variable is y, then you can either type

logit y cs_*, or noconst

to estimate parameters corresponding to the odds at the reference points. 
Alternatively, you can type

gene byte baseline=1
logit y baseline cs_1-cs_4 cs_6-cs_10, or noconst

and the parameter corresponding to the variable baseline will be the value 
of the odds at the reference point corresponding to cs_5, and the 
parameters corresponding to cs_1-cs_4 and to cs_6-cs_10 will be odds ratios 
between the odds at the other reference point and the odds at the ommitted 
reference point. Note that the ommitted reference point must be in the 
completeness range of the spline.

I am currently preparing a paper on reference splines for submission to 
Statistics in Medicine, explaining these issues. I can send you a copy of 
the current preliminary draft on request.

I hope this helps.

Best wishes

Roger


--
Roger Newson
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
United Kingdom

Tel: 020 7848 6648 International +44 20 7848 6648
Fax: 020 7848 6620 International +44 20 7848 6620
  or 020 7848 6605 International +44 20 7848 6605
Email: roger.newson@kcl.ac.uk
Website: http://www.kcl-phs.org.uk/rogernewson

Opinions expressed are those of the author, not the institution.

*
*   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/