I'm trying to run a linear regression analysis to identify a rather obvious inverse correlation between two variables over time. My problem is that I only have 6 timepoints, but I do have 3 repeated measures for each timepoint.
A simple linear regression using STATA including all repeat measures gives a p value of 0.005. This is clearly not valid because the repeat measures are almost (but not exactly) the same mesurements taken at the same timepoint which is artificially increasing my sample size. Looking at each repeat seperately doesn't show significance.
Is there a way of weighting to account for multiple sampling (and different variances). Could I try a regression with ANOVA (am I thinking along the right lines).
This is not my field so any help in simple(ish) language would be greately appreciated. If you can suggest specific STATA code that would be even better.