Usage ----- ^genvars^ , varA(numlist) varB(numlist) ... varX(numlist)
Description ----------- ^genvars^ generates new variables in a recursive fashion. This is best illustrated by example:
.^genvars^ age(0(10)100)
generates a new variable age that has observations 0,10,20..100.
.^genvars^ age(0(10)100) gender(1,2)
generates a new variable age that has observations 0,0,10,10,20,20..100,100 and for each age the new variable gender has values 1 or 2.
.^genvars^ age(0(10)100) gender(1,2) smoking(1,2,3)
Remarks ------- ^genvars^ was written for use with regression modelling. Sometimes your orginal data may miss particular ages. By generating new covariate patterns you can use the generated data and -predict- to view your model. See Examples.
Examples --------
.use "astma file" .frencurv , xvar(age) power(3) gen(agesp) .logit astma agesp* gender
Now - your model doesn't contain any data for ages 46, 48, which leaves holes in your graph. Then do:
.drop _all .genvars age(0(1)100) gender(1,2) .frencurv , xvar(age) power(3) gen(agesp) .predict linpred, xb .sort age gender .generate refcat=linpred[1] .replace linpred=linpred-refcat .generate or=exp(linpred) .graph linpred age if gender==1
or use the utility @pspline@ which does the same.
Author ------ Jan Brogger, jan.brogger@@med.uib.no
Shameless plug :) -------------- I am looking for internship & residency in internal medicine/pulmonary medicine US/EU. Am doing PhD in respiratory epidemiology now.