Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: interpreting multivariable fractional polynomials


From   "Casey P. Durand" <durandca@usc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: interpreting multivariable fractional polynomials
Date   Thu, 13 May 2010 12:17:15 -0700

Hi folks,

I'm new to using multivariable fractional polynomials -mfp- in Stata,
and I have a question about interpretation of the dual coefficients
from a second degree FP .  The best fitting model was a (-2, -2)
transformation, and this is my output from the logistic regression
using those transformations (street_1 & street_2):

Logistic regression                                   Number of obs
=        364
                                                               LR
chi2(11)     =      79.55
                                                               Prob >
chi2     =     0.0000
Log likelihood = -113.04482                       Pseudo R2       =     0.2603

------------------------------------------------------------------------------
   meets_rec |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    street_1 |   1.002819   .9828367     1.02   0.308    -.9235053    2.929144
    street_2 |  -15.11272   5.243973    -2.88   0.004    -25.39072   -4.834722
child_gend |  -1.904081   .4017071    -4.74   0.000    -2.691413    -1.11675
             |
 child_ra |
          2  |   .0318764   1.010836     0.03   0.975    -1.949325    2.013078
          3  |    .136247   .4551104     0.30   0.765    -.7557529    1.028247
          4  |   .0477573   .7104908     0.07   0.946    -1.344779    1.440294
          5  |   -1.37826     .71353    -1.93   0.053    -2.776753    .0202333
          6  |   .4384325   .6696987     0.65   0.513    -.8741529    1.751018
             |
 free_lunc |   .9778007   .3936116     2.48   0.013     .2063361    1.749265
   child_age |  -.8609912   .1538146    -5.60   0.000    -1.162462   -.5595202
    preserve |  -.6135725   .4557935    -1.35   0.178    -1.506911    .2797663
       _cons |    12.0511   2.034686     5.92   0.000     8.063188    16.03901
------------------------------------------------------------------------------

Based on what I've seen in other published papers, it seems that I can
present the results graphically (i.e. predicted odds v. the
untransformed variable), and/or present odds ratios using hypothetical
values.  My question is how to do the latter.  Given the range of
possible answers on the continuous, untransformed "street" variable
(1-5), I'd like to give the OR predicted by the above coefficients
using a score of 1  as the referent category, while holding all other
variables at their mean, for the scores of  2,3,4 & 5.  Though I've
read the help section and manual  for -adjust- and -margins-, I can't
figure out how to get the predicted odds ratios and 95% CIs for these
hypothetical scores.

Thanks in advance for any help you can provide,

Casey
*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index