predict.BTmto estimate abilities with non-player abilities set to non-zero values (for models with a fixed reference category).
qvcalc.BTabilitiesmoved over from package qvcalc.
predict.glmmPQLis 0 if a fixed effects model has been fitted, 1 otherwise.
BTabilities now works (again) for models where the reference category is not the first player. Players are kept in their original order (levels of
player2), but the abilities are returned with the appropriate reference.
BTabilities now works when ability is modelled by covariates and some parameters are inestimable (e.g. as in
predict.BTglmmPQL now works for models with inestimable parameters
NAfor unidentified abilities
player2factors. Also handle unidentified coefficients correctly.
BTglmmPQLto avoid conflict with lme4 (which loads MASS).
BTmso that it is able to find variables when called inside another function (stackoverflow.com question 14911525).
anova.BTmlist to work for models with random effects
allow models to be specified with no fixed effects
offset argument to work as documented
corrected documentation for
predict.BTmnow works for models with no random effects and handles new individuals with missing values in predictors.
BTm.setupcausing problems in finding variables when
BTmnested within another function.