Comparison of alternative models for predicting LAI in moist forest and miombo woodland of Hanang District in Tanzania.

Predictor sets include: structural variables only; environmental variables only; structural and environmental variables combined (see Table 1). Results are for models reduced by stepwise selection (see S1 and S2 Tables for details on global models and covariate estimates, respectively). Statistics: D2, percent deviance explained; ΔAIC, change in Akaike Information Criterion compared with null model; LRT, likelihood ratio test comparing final models with their respective global models at P ≤ 0.05; MSEP, mean square error of prediction; W, Wilcoxon Mann-Whitney statistic used to estimate prediction bias.Comparison of alternative models for predicting LAI in moist forest and miombo woodland of Hanang District in Tanzania.