How to combine variables?

[CactusPete]

I am running a proc surveymeans with the resulting table attached. I am using BPX (blood pressure) laboratory data and there's multiple recordings (BPXSY1, BPXSY2,...) so I am trying to average them. I was able to get an average but the values it used were not weighted and did not match the values in the table attached. Is there anyway of doing this?

[ballardw · Accepted solution]

It is a good idea to post links to related questions.

You would have to show the code that you used for "averaging"  and then generating BOTH tables that say "don't match".

And tell us which values "don't match". I would not expect the confidence intervals or errors to match. That is why we have the survey procs to handle complex weighting. If you use Proc means on the same data you will not get the same intervals and depending on the actual values and weights there can be notable differences.

If by "mean" you meant the per observation mean of the three variables and then asking for the overall mean with this procedure, you don't show the additional variables that I would expect for that, just your "raw" variables.