Dear all,

I am struggling to find the best way to analyse my dataset. I have to compare fish diversity in restored and unrestored river sites over time. Fish samples were collected from 80 comparable rivers. However, in some rivers the sampling was conducted only 1 time, in other rivers the sampling was conducted up to 15 times (once per year).

To start I thought to use this simple mixed effect model:

Fish diversity~restoration*years from restoration, random=(1|rivers).

However, I believe I have to run my model accounting for temporal autocorrelation (ARMA correction in lme) as data in many rivers are collected mulitple times over years.

My question is: can I use such a correction when in my daset I have rivers in which the sampling was conducted only 1 year while in other rivers the sampling was conducted multiple years (up to 15)?

Thanks for your consideration

Alessandro