Hello Stata users,

I am performing a cosinor analysis on seasonal iodine dispersion. To do this, I have appended 2 survey data sets, with two different probability weights totaling about 3000 observations. I understand that given missing data, etc., the final regression will not contain ~3000 observations, but this number changes significantly when I include the probability weights. When I use one, I get an output with 888 observations, and when I use the other, my output suddenly has close to 2000.

My question here is how do I justify the use of one pweight over the other? Is there a way to include both in my regression?


Thank you for your time!

-Hannah