General stata resources for analyzing J or U curve shaped data

[Al Bothwell]

I am analyzing data from a longitudinal data on hypertension. The outcome is all-cause mortality (yes or no). The literature says that systolic blood pressure reduction has benefit up to a low point (nadir), further reductions beyond the low point are harmful.

So the data will be either J or U shaped (positive effects of blood pressure reduction up to a point and thereafter negative effects of blood pressure reduction).

I would appreciate any Stata resources (code, textbooks, discussions from this forum, commands) on how to analyze this type of data. Right now, I am thinking of doing a type of non-linear logistic regression, I could be wrong.

Thank you,
Al Bothwell

[Carlo Lazzaro]

Al:
you may want to consider something along the following lines:

xtlogit mortality c.SBP##c.SBP <otherpredictors> <controls>, re

The interaction includes both a linear and a square term for systolic blood pressure (SBP) in order to investigate the presence of a turning point when adjusted for the remaining predictors.
In addition, you may want to consider a clustered standard error (on -panelid-) provided that the number of your panels is large enough.

[Dario Maimone Ansaldo Patti]

You may want to look at the user-written command utest. If I remember correctly, it works with logit models too.

ssc install utest

[Al Bothwell]

Thank you Carlo and Dario.