topic Re: Censored normal regression to adjust for hypertension treatment effect in SAS Programming

Censored normal regression to adjust for hypertension treatment effect

LucyLou — Mon, 14 Oct 2013 21:12:18 GMT

I am trying to figure out how to run a censored normal regression in SAS. I have cross-sectional data that I am analyzing to examine the association of an independent variable with systolic blood pressure (SBP). The observed SBP in treated individuals (HTNMed=1) is right censored, therefore I have been told to use censored normal regression. I have found how to do this in STATA (cnreg), but I can't seem to figure it out in SAS.

Does anyone know how to do this?

Thank you!

Amanda

Re: Censored normal regression to adjust for hypertension treatment effect

SteveDenham — Tue, 15 Oct 2013 12:29:05 GMT

Do you have SAS/ETS licensed? If so, take a look at PROC QLIM.or PROC HPQLIM. I don't think there is a good way to approach this with SAS/STAT, but you might consider PROC LIFEREG for tobit regression--check Example 51.2 Computing Predicted Values for a Tobit Model, and see if you can substitute your values into the example there.

Steve Denham

Re: Censored normal regression to adjust for hypertension treatment effect

LucyLou — Tue, 15 Oct 2013 17:12:15 GMT

Thank you! It looks like PROC LIFEREG will do the trick.

Re: Censored normal regression to adjust for hypertension treatment effect

LucyLou — Sat, 01 Feb 2014 23:11:23 GMT

Hi again,

I have been successful in using Proc Lifereg to run censored normal regression for my analysis (thank you!). Now I am trying to predict mean outcome values (blood pressure) for 4 different models starting with a minimally adjusted model to one adjusting for all covariates, and am having trouble figuring out how to do this.

I tried to follow this example: http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_lifereg_sect035.htm, but I'm starting to think that this is not the answer to my problem because my mean blood pressure is not changing more than a few tenths when I move from model 1 to model 4. I know this cannot be the case based on the results of the Proc Lifereg.

Does anyone know what I may be doing wrong or how to calculate mean outcome values for tobit models with right censoring?

Thank you in advance for any advice you have.

Amanda