Need help with unbalanced design

Hi Everyone,

I have data from an experiment with a RCBD. The main independent variable of interest is crop variety (var). In 2011, we conducted the experiment in one location and there were four different crop varieties. In 2012, we conducted the experiment in two locations and added a fifth crop variety. So the design is unbalanced because the 2011 data lack the fifth crop variety.

I am using proc mixed because I was under the impression that this procedure could deal with missing data and unbalanced designs. When I run the model I get output that looks fine, but when I look at the log there is a message the the G-matrix is not positive definite.

Can anyone please provide some advice about how to cope with unbalanced designs in proc mixed.

Thanks!

Gigi