Parametric survival analysis of the timing of leaf beetle and rust attack.

Beetle and rust data were analysed separately, with attack treated as “mortality”. The lines show the null model (with a single distribution location and scale parameter) and the change in log likelihood as terms for population (Emosson or La Fouly), treatment (seven levels), and the population by treatment interaction were sequentially added. The final three columns provide likelihood ratio tests of the significance of each term.