I have been asked by various StataListers to add certain options to the
-oddsrisk-
program that I posted to SSC in December. I have submitted the revised
command
and help file to Kit for re-posting.
The new oddsrisk program now has -if- and -in- capabilities and allows
frequency weights.
This is important if one wishes to estimate odds and risk ratios directly
from a 2 x 2
table. Of course, Stata has a several commands that can be used to calculate
risk
ratios and CIs (using Woolfe's method) from a 2x2 table; eg -csi-. The
option has been
added to allow users to model estimated risk ratios from both full and
tabulated data.
For example, given a 2x2 table of -death- (response) and -anterior- (risk
factor or exposure),
you would model death on anterior as:
oddsrisk death anterior [fw=count]
A listing of odds ratio, estimated risk ratio, and estimated 95% risk ratio
CI's are displayed.
If oddsrisk is used on the complete data set, the command would be
oddsrisk death anterior
without the need for a frequency weight of counts.
Care must be taken when using -oddsrisk- to estimate relative risk ratios
for a number of
predictors; i.e., a risk factor and confounders. There is evidence that
confidence intervals are
biased. However, the bias appears to be relatively minimal and the values do
not vary
appreciably from CIs from a Poisson with robust SE or log-binomial. All
three methods are
close approximations of the risk ratio and corresponding CIs. References are
given in the help
file.
Joseph Hilbe
**************Biggest Grammy Award surprises of all time on AOL Music.
(http://music.aol.com/grammys/pictures/never-won-a-grammy?NCID=aolcmp003000000025
48)
*
* 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/
