# How to model binary dependent data with temporal autocorrelation?

I am trying to model annual tree nut production using climate predictors.

The nut data (dependent) is a binary timeseries (0,1 - representing unsuccessful and successful nut production), with one observation per year, and with 90 years of data and two missing years (88 onservations).

The independent variables are monthly climate variables, including months in previous years (for example, Temp.July.t, Temp.July.t-1)

I'm using R, and have an basic-intermediate knowledge of statistics.

My problem is that the dependent data has strong temporal autocorrelation (nut production cannot be successful two years running). I'm looking for a pointer towards a technique that will allow me to deal with the autocorrelation in the binary data and create a statistical model that allows me to investigate the relationship between nut production and climate.

Thank you.

• Are you saying there's a strong negative autocorrelation, ie that if year t=1, year t+1 must equal 0? Is it possible to have two 0's in a row? Oct 17, 2012 at 15:13
• It is possible to have two 0's in row, but not two 1's. So yes, if t=1 then t+1 = 0, but not necessarily the other way around (so if t=0, t+1 can be either 0 or 1. Oct 17, 2012 at 15:28