Jisheng is having trouble posting this:
m.p.
Dear all,
For the attached data set, I fitted a censored regression, where
treatment=1 represent SBP is not observed, but actually >= the censored
(listed) value. Following commands are used to calculate the residual,
for which I hope that on average they are less than 0 for individuals
with treatment==1, because the observed (listed) value is rightly
censored and are expected to be less than the true value, and hence,
less than the predicted value if the model is OK. But, I obtained a mean
of 3.55, which is significantly higher than 0. WHY?
With thanks,
Jisheng.
. cnreg sbp age, censor(treatment)
. predict y_sbp
. gen r_sbp = sbp - y_sbp
. sum r_sbp if treatment==1
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
r_sbp | 126 3.549016 16.92667 -40.63338 46.86662
Attachment:
temp.dta
Description: Binary data
