Trend analysis for non-longitudinal data

[lizzy28]

Hi everyone, 

I have three year data with the same measurements per observation, and there are 4000-6000 observations in the data set of each year. The three-year data is not longitudinal as there are only five observations that are in the data of all the three years. Is it possible to run any trend analysis this case (rather than simple description)? The sample looks like below:

Year	Year1	Year2	Year3
N	6,000	4,000	5,000

flag_Year1	flag_Year2	flag_Year3	N
1	0	0	3,000
1	0	1	1,500
1	1	0	1,495
1	1	1	5

Thank you!

Lizi

[jk123xyz]

The simple answer is just to make year (or time) a continuous independent variable.  You can code year as 1, 2 or 3.

proc reg; model outcome = year; run;

This will give you a slope for the trend across time.