12-02-2016 02:26 AM - last edited on 12-02-2016 05:49 AM by Reeza
Look at the association between stroke and high blood pressure by sex. Do you see any signs of confounding or interaction/effect modification? Regardless of your conclusion, calculate the Cochran Mantel-Haenszel prevalence rate ratio and compare it to the crude prevalence rate ratio. Is confounding present?
When I am using SAS, I get an error of 50% of cells have expected counts less than 5 error message. How can I correct this? I'm trying to get the relative risks 1 and 2 and I am unable to get them. Can you help me with this? Please give me some specific code if you can because I am stumped.
12-02-2016 10:38 AM
Does the text of the message say ERROR or WARNING?
SAS provides warnings for some procedures to let you know that what you are doing may either be inappropriate or that, in this case, the sample distribution, may have issues in interpretation.
What exact approach, code if you have it, did you use?
I would guess that you dumped the data into proc freq and that the blood pressure, as a continuous measurement, had a very large number of categories. If so you have a number of approaches. One might be to create a categorical variable for blood pressure with 2 values that mean "high" and "not high".
Some example data of what you have, enough to illustrate your data question, is usually appropriate.