Count data over years and repeated measures

[mthorne]

We have done long-term weed control trials over three to six years using density counts each year to measure efficacy of the treatments. Each year a new group of stems from underground rhizomes emerge in each plot. I am not sure that the counts each year are a true repeated measure because we are not measuring (counting) the same stems. If it were length of growth of the same stem over time, then I could see that as a repeated measure, but we are counting a new crop of stems each year. Does anyone have any thoughts on this?   

[StatDave · Accepted solution]

If you consider the plots as the units being measured, not the individual plants, and you are simply counting the number of emergent plants in each plot at each time point, then this sounds like repeated measures count data that could be modeled with, for instance, a GEE model. In PROC GEE you would specify an appropriate response distribution like Poisson or negative binomial in the DIST= option in the MODEL statement. In the REPEATED, specify the plot identifier in the SUBJECT= option in the REPEATED statement and it never hurts to also specify the time point variable in the WITHIN= option.