Version 1.9
-----------------------------------------------------------------------------
General:
* Fixed many issues related to the latest update of ggplot2.
* Argument `show.se` is now deprecated. Use `show.ci` instead.
* Redesign of computation of frequency tables for `sjp.frq` and `sjt.frq`, being more robust and generally working with labelled, non-labelled, numeric, character vectors and factors.
* Redesign of computation of frequency tables for `sjp.grpfrq`, `sjp.xtab` and `sjt.xtab`, being more robust and generally working with labelled, non-labelled, numeric, character vectors and factors.
* Better automatic handling of variable and value labels that are used for labelling plot axes and titles or table columns.

Changes to functions due to new ggplot2-version:
* `sjp.grpfrq` no longer has plot type `type = "histogram"`. Maybe re-implemented in a later update. Due to this change, arguments like `showMeanIntercept` and similar were removed.
* Plotting functions no longer have argument `labelPosition`. Instead, use arguments `vjust` and `hjust`, which correspond to the same ggplot2-aesthetics according to their possible values.

Changes to functions:
* `sjp.lm` gets a `group.estimates` argument to group estimates in forest plots and colour them according to group assignment. Use arguments `show.legend` and `legendTitle` to modify group legend.
* `sjp.lmer` and `sjp.glmer` can now plot random effect parts of random slope-intercept models (with `type = "rs.ri"`), where regression lines or predicted probabilities of random intercept and slopes are plotted.
* Intercept line plotting in `sjp.int` for `type = "cond"` was removed.
* Line geoms for `type = "cond"` in `sjp.int` now always start at y-position zero, to better indicate the effective change of interaction effect compared to base reference. Now, the y-position indicates the change in the reponse due to the interaction effect.
* `sjp.int` gets a `geom.size` argument to specify line width.

Bug fixes:
* Argument `ci.hyphen` in function `sjt.lm` and `sjt.lmer` was not correctly applied to confidence intervals of standardized beta values.