Missing values generated when used command predict

[Seungwoo Kang]

I am tried to produce sensitivity and specificity by manually through predict using the result of svy:logit. Below is the syntax I wanted to apply.

svy:logit y x1 x2 ... , nolog or allbaselevels
predict prediction, pr
gen byte predicted = (prediction > 0.1) & !missing(prediction)
svy, subpop(if srh2021 == 1): proportion predicted
svy, subpop(if srh2021 == 0): proportion predicted

However, it shows

predict prediction, pr
(999 missing values generated)

thus, missing portion deterred having an expected result. How to cope with those 999 missing values?

[Andrew Musau]

You cannot predict values for observations that have missing entries. See

help e(sample)

to determine what observations constitute your sample and read up on listwise deletion.

[Carlo Lazzaro]

Our Colleague Seungwoo Kang posted a very similar query at https://www.statalist.org/forums/for...ommand-predict.
Multiple posts on (more or less) the same topic started by the same poster increase the risk of scattering potentially useful replies in different messages, which is neither useful for the original poster, nor for those interested in following the query.