Dear Lara, you introduce random effects if you have grouped or repeated-measures data, to account for the resulting covariance/non-independence, or if you are interested in modelling variability at different hierarchical levels. From your description at least, all your data points seem independent of each other. qinit is a covariate, i.e.: you probably want to model this as: glm(clutchsize~treatment*species+qinit) HTH, Yann
Dr. Yann Clough Agroecology Department of Crop Sciences G?ttingen University Grisebachstrasse 6 Room 3.131 (3rd floor) D-37077 G?ttingen (Germany) Tel.: (+49) 0551 3922358 Fax.:(+49) 0551 398806