Apply survey weights to calculate significance of trend over time

[he2182]

Hi,

I'm using publicly available data to look at trends of illicit drug use of a ten year period. I was trying to use the cochrane armitage trend test but it's not available in proc surveyfreq. Two questions:

1) How do you account for survey weights when calculating trends across time?

2) The survey weights are person level but the data is collapsed count data by yeat and illicit drug use (y/n).  How do I recalculate the survey provided person level survey weights?

Any help is appreciated! Thank you.

[ballardw]

I would take a look at SURVEYREG with year as your independent variable. Tests on the slope would aim towards significanct.

Also look at DOMAIN analysis for sub-population comparison

[he2182]

Thanks. Proc Surveyreg would work if my dependent variable was continuous but I'm look at a categorical (Yes vs No) outcome of past month illicit drug use.

I'll look into Domain analyses.