Recently, a new version (0.9) was uploaded to CRAN, and should be available now to all users. Several improvements and changes were made. Some of these changes may affect existing users. Therefore, we provide an overview of the improvements and changes to influence.ME:

• To better align with R terminology, the estex() function was renamed to influence().
• Several of the existing functions were rewritten, so that they are methods to R generic functions. ME.dfbetas() and ME.cook() were renamed to dfbetas() and cooks.distance() and all plotting functions were removed from the package and replaced by a plot() function.
• The plot=TRUE parameter in the cooks.distance() and dfbetas() functions are is no longer available: plots should be called for using the plot() function.
• In addition to these changes, influence.ME now provides several new features:
  • The sigtest() function allows users to test whether the level of statistical significance of a parameter estimate is affected by the presence of influential data.
  • Users can now also test wether lower-level observations affect the multilevel model outcomes (rather than only evaluating the influence of nested groups of cases).
  • Plots on dfbetas / sigtest / pchange can now plot values values that exceed a cutoff value visually distinct.

As a result of some of these changes, users may need to modify their code (slightly). The new structure of the influence.ME package is much more in line with R standards. Remember that the version number (0.9) indicates that influence.ME is still in beta stages of development, and that changes may take place in its design. Please feel free to contact me with any questions, comments, and / or problems you may have regarding our software.

I’m nowhere near finished, but I think it is already looking good. I would love to receive some feedback from regular readers. What do you think: is this going in the right direction, or was the previous layout of Curving Normality better?