The basic code will look like this: proc glimmix data=yourdata method=quad(qpoints=4); effect splTime = spline(Time / knotmethod=equal(2)); model Survival = splTime; output out=smoothed pred=smooth_survival; run; You may need to adjust the Number of knots (1 might be enough).   However, this won't contrain predicted survival to 1.0 at Time=0. I thought this could be done with the Parms statement, but I was mistaken. Instead, you could model the compliment of survival with no intercept, and that should do the trick.   data yourdata2; set yourdata; csurv = 1 - survival; run ; proc glimmix data=yourdata2 method=quad(qpoints=4); effect splTime = spline(Time / knotmethod=equal(2)); model csurv = splTime / noint; output out=smoothed pred=smooth_csurv; run; Then take the compliment of csurv to get back to the smoothed survival estimates you want. I haven't ever tried to do quite what you are doing, as we always present the standard Kaplan-Meier curve. You should consider smoothing the upper and lower confidence bounds as well, to convey the uncertainty that goes along with this curve. ... View more

Interesting idea, but how do I test it?   I've attached the file.   Norman. ... View more

I recommend writing a very small program, using either a SAS supplied data set or build one with a data step, and minimal code to provide an example of what happens with your system.   While it shouldn't be part of the problem, all of the style overrides and macro variables and no date make executing your code problematical for testing ... View more

Bigger is not better.   The more variables that are included, the better the model fit (and the lower the deviance). However, the idea is to construct a model with only as many variables as are required.   Removing variables will decrease the model fit, but this, in itself, is not a bad thing.   Norman. ... View more

No. You can't . It is high dimension data. There are lots of  intercept and slope satisfied multiple linear regression Y=b0+b1x1+b2x2+b3x3 But only one of them is nearest center(has the smallest NORM), and SAS use OLS or ML to get that one . ... View more

It is possible. It usually means that you have many explanatory effects compared to the number of observations. I can't remember the rule-of-thumb right now, but some experts recommend a minimum of 10-20 observations per regressor.  If you have only 100 observations and you try to use 75-100 effects, your model will be bad and you might get a negative adjusted R-squared value.   Limit yourself to main effects instead of many interactions, or choose fewer explanatory variables. ... View more

If you had SAS/ETS, you could use a single call to PROC SYSLIN.   Although I am not an expert in this area, I believe you can call PROC REG twice to perform this analysis, based on comments in Using SAS for Econometrics, which is the SAS supplement to Hill, Griffiths and Lim, Principles of Econometrics, 4th ed.  Here is a link to the Chapter 10 exercises, which compare 2-stage PROC REG with PROC SYSLIN.   From the Chapter 10 code, it looks like you use PROC REG to generate residuals for the first stage, and you call PROC REG again and use those residuals in the second stage regression to get Hausman's test.  ... View more