Data for this study were collected by the US Forest Service and the Pennsylvania Bureau of Forestry on nearly 600 plots in central Pennsylvania. Tree and stand characteristics recorded between 1978 and 1985 include estimates of percent defoliation on individual trees.

Logistic regression using maximum likelihood estimation was employed to model individual-tree mortality of 15 species in central Pennsylvania that had been defoliated by the gypsy moth. Defoliation was estimated to the nearest ten percent for individual trees. Other variables used for prediction included stand basal area and an individual-tree relative basal area index. Success ranged from no fit for three of the species to an R value (a derivation of Akaike's information criterion) of .613 for white oak. The inclusion of defoliation in the models had a varied effect. For four of the species percent defoliation was not significant. For hickory and white oak respectively, percent defoliation raised the R value by .305 and .290 percentage points. As many as five models for each species were developed: one or two models with no defoliation measure in the model and one each for one, two or three consecutive years of defoliation measures.

A beta and gamma function were used to model individual· tree basal area growth for the same 15 species. The models were fit using nonlinear least squares. Variables used include the relative basal area index, stand basal area, site index and a defoliation index that incorporated three years of individual-tree, percent defoliation. The beta and gamma functions fit equally well with values of (1 - relative mean square error) ranging from .1967 to .6290. Results for both models are presented for each species.

The defoliation index was a significant variable for five of the fifteen species: white, chestnut, red, and black oak and sassafras.