I have just tried using the predxcat command, but I noticed that the
confidence intervals seem too narrow and are pretty much the same as the
CI's obtained when using:
Adjust x2 x3 x4, by(x1) pr ci
These CI's are calculated using the standard error from the predicted index,
not the predicted probability---correct?
Christer wrote:
Check out
Predxcon or predxcat
or
logpred , regpred
or
The commands at Scott Long's Homepage (http://www.indiana.edu/~jslsoc/)
Greg wrote:
To predict probabilities and 95% CIs of the predicted probabilities after a
multivariate logistic model, I have previously used the following predict
command:
logistic y x1 x2 x3
predict lr_index, index
predict se_index, stdp
gen p_hat = exp(lr_index)/(1 +exp(lr_index))
gen lb = lr_index - invnorm(0.975)*se_index
gen ub = lr_index + invnorm(0.975)*se_index
gen plb = exp(lb)/(1+exp(lb))
gen pub = exp(ub)/(1+exp(ub))
bysort x1: tabstat p_hat plb pub
However, this code does not allow me to predict probabilities,
by(x1),�while keeping the other covariates held constant, as the adjust
command does.
*
* 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/
