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

Re: st: Missing outcome variables - how to deal with these? OR moels and reality

From   "jmpsouza" <>
To   <>
Subject   Re: st: Missing outcome variables - how to deal with these? OR moels and reality
Date   Sun, 24 May 2009 14:23:04 -0300

Dear people:
may I present a situation very usual, indeed one that already happened to me, a possible way to handle it and the question whether the conduct is reasonable:
1. the data come from official registry;
2. the outcome is birth weigth, coded 0 for normal and 1 for low, no missing; 3. the model is logistic (just for simplify, since it was used glm poisson robust as an estimate of the prevalence ratios); 4. var1, var2, var3, var4 with very very few missings (let´s say less than 2%);
5. var5, with 4 categories, but with about 20% missing answers;
6. var6, same as var5;
7. var5 received then 5 categories: 1 , 2 , 3 , 4, the same original ones and 5, when missing;
8. tab(var5), gen(var5);
9. same for var6;
10. any of var51 or var52 or var53 or var54 is used as reference, the others, including var55 are in the command ;
11. same for var 61, var62 etc.
12. the command is logistic weigth var1 var2 var3 var4 var52 var53 var54 var55 var62 var63 var64 var65; 13. in the analysis var55 and var65 are placeholders, but their odds ratios may be useful. The question: may this be used, is it too difficult to accept? Is it a situation for using -ice-?
Excuse me for so long post.
JNosé Maria

Jose Maria Pacheco de Souza, Professor Titular (aposentado)
Departamento de Epidemiologia/Faculdade de Saude Publica, USP
Av. Dr. Arnaldo, 715
01246-904  -  S. Paulo/SP - Brasil
fones (11)3061-7747; (11)3768-8612
fax (11)3714-2403
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index