Programming the statistical procedures from SAS

Proc Panel with randomized effects + fixed effects

Reply
Occasional Contributor
Posts: 8

Proc Panel with randomized effects + fixed effects

Hello,
I am a relatively new SAS user, just starting out with modeling so forgive my possible ignorance.

I have some time series data where some data changes over time, while others do not, however they are different across individuals (in this case, lets say height and weight stay the same but other biological variables change throughout the day).

It has been suggested to use proc panel with randomized effects, because fixed effects reports errors as not all change across time. Therefore, the following procedure was run:

proc sort data=body2;
by person stage;
run;

proc panel data=body;
id person stage;
model vo= female1 stage wt ht bodyfat meanamp tp hr rpe / rantwo;
run;


I believe rantwo should be used as there is both a cross-sectional and time-series component? If someone can correct/affirm this for me, that would be great.
After running this procedure, we determine the coefficients for the variables which do not change over time and then restrict them, while running a fixed-effects proc panel model as follows:

proc sort data=body2;
by person stage;
run;

proc panel data=body2;
id person stage;
model vo= female1 stage wt ht bodyfat meanamp tp hr rpe / fixone;
restrict female1 = -1.1803, wt=0.29765, ht=-0.81169, bodyfat=-0.59994;
run;

First, can anyone advise whether this is proper modeling technique? Is there a better method/rationale? Second, when this is run, all the variables restrict fine except for the variable female1, which is a dummy (1,0) variable. This restricts under the "restrict0" variable which shows up when coefficients are restricted, not under the female1 variable, which still shows up as a 0. I'm not sure how to describe it better, but can anyone explain why this happens?

As I said, I'm open to other thoughts as to how to properly run this model. Thanks!!
Ask a Question
Discussion stats
  • 0 replies
  • 58 views
  • 0 likes
  • 1 in conversation