# st: svy and mlogit with Stata9.2

 From "horace Roman" To statalist@hsphsun2.harvard.edu Subject st: svy and mlogit with Stata9.2 Date Wed, 28 Feb 2007 14:25:32 -0500

```Hi,

I have a big problem with survey commands
I'm working on Stata9.2; I'm studying the likelihood to have an 3
classes outcome (clPT) depending on several risk factors (momsmoke,
momdrink, etc)

I used the help survey where I found a list of commands used in the analysis

however, when I try to do them (clPt is my qualitative variable) I obtain

. svy: proportion  clPT
(running proportion on estimation sample)
Mata run-time error
r(3103);

moreover, I need to do a multinomial regression model; when I'm doing
like it is shown in the commands list I obtain

svy: mlogit clPT momsmoke  momdrink, rrr
(running mlogit on estimation sample)
Mata run-time error
r(3103);

however whether I writte differently it seems to work

. svymlogit clPT momsmoke  momdrink, rrr

Survey multinomial logistic regression

pweight:  pw                                      Number of obs    =     65367
Strata:   <one>                                   Number of strata =         1
PSU:      <observations>                          Number of PSUs   =     65367
FPC:      fpc                                     Population size  = 2631930.9
F(   4,  65363)  =     48.69
Prob > F         =    0.0000

------------------------------------------------------------------------------
clPT |                   RRR   Std. Err.      t    P>|t|
[95% Conf. Interval]
-------------+----------------------------------------------------------------
med preterm  |
momsmoke |     .7021902   .0179722   -13.81   0.000     .6678336    .7383143
momdrink |      1.17521   .1111002     1.71   0.088     .9764373    1.414446
-------------+----------------------------------------------------------------
very preterm |
momsmoke |   .8528374   .0313991    -4.32   0.000     .7934632    .9166545
momdrink |      .9132249   .1123855    -0.74   0.461     .7175021
1.162338
------------------------------------------------------------------------------
(Outcome clPT==term is the comparison group)
Finite population correction (FPC) assumes simple random sampling without
replacement of PSUs within each stratum with no subsampling within PSUs.

Horace

--
Horace Roman

Ashley House, Condo #8G
14 Lockwood Drive
29401 Charleston,
South Carolina, USA
Tél : 1- 843-723-9630
Fax 33 - 825 175 480
Office: 1-843-792-6315

Gynécologue-obstétricien
Clinique Gynécologique et Obstétricale
CHU Charles Nicolle
76031 Rouen

Unité de recherche épidémiologique en santé périnatale et en santé des
femmes, INSERM U149,
82 Avenue Denfert-Rochereau,
75014 Paris
```

Attachment: Hi statalist.doc
Description: MS-Word document