BTm finds variables passed to outcome, player1 etc, so that it works when run in a separate environment.anova.BTm now respects test and dispersion arguments for models that inherit from glm.anova.BTmlist affecting models where ability is modelled by predictors but ability is estimated separately for some players due to missing values.glmmPQL affecting models with . in formula and either offset or weights specified.Diff() that gave warning under R-devel.if statements where argument could be > 1.qvcalc.BTabilitiespredict.BTm to estimate abilities with non-player abilities set to non-zero values (for models with a fixed reference category).qvcalc.BTabilities moved over from package qvcalc.level in predict.BTm and predict.glmmPQL is 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 player1 and 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 chameleons.model on ?chameleons).
predict.BTglmmPQL now works for models with inestimable parameters
BTabilities now returns NA for unidentified abilitiesplayer1 and player2 factors. Also handle unidentified coefficients correctly.glmmPQL object BTglmmPQL to avoid conflict with lme4 (which loads MASS).BTm so that it is able to find variables when called inside another function (stackoverflow.com question 14911525).fixed anova.BTmlist to work for models with random effects
allow models to be specified with no fixed effects
fixed offset argument to work as documented
corrected documentation for citations data
predict.BTm now works for models with no random effects and handles new individuals with missing values in predictors.BTm.setup causing problems in finding variables when BTm nested within another function.