I tried the command "aflogit" to obtain adjusted Population Attributable
Fractions (PAFs) after logistic regression (command "logistic").
I have 3 questions:
1. The command seemed not to work properly. I carefully followed the
instruction but I received the error message "l_1 not found". I was
wondering whether the command still works on Stata 8.0
Anyway, the error message appeared after the PAF for the variable I was
interested in had been displayed. Also, the command worked fine when no
dummy variables were used (i.e., the "xi" function was not used)
2. If I am correct, the command simply:
a. gets the OR from the estimation produced by "logistic" (in my case
OR=3.7)
b. then calculates the AF in the exposed subjects (EAF), i.e.,
(3.7-1)/3.7=0.73
c. Multiplies the EAF for the prevalence of the exposure among cases (0.29
in my case). i.e., 0.73*0.29=0.21
Is that correct?
3. The help file says that the command calculates (function "cc") PAFs from
data that come from an unmatched case-control study. Why "Unmatched"? Does
that simply mean that the command do not work with "clogit" as I understant
from the help (therefore, matched refers to 1:1 matching?)?
My case-control study has a frequency matching. Is it not correct to use the
command (and the formula above at point 2) when a matching is present?
Thanks a lot,
Andrea
Andrea Baccarelli, MD, MPH, PhD
Genetic Epidemiology Branch
Division of Cancer Epidemiology and Genetics
National Cancer Institute, NIH, DHHS
6120 Executive Blvd. - EPS 7110
Bethesda, MD 20892-7236
Tel. (301)-496-5786
Fax (301)-402-4489 [email protected]
