"normality" and "compositions" packages for Red-R

Hi,

I recently updated two packages: "normality" and "compositions".

"normality" contains some widgets related to normal distributions. I included several statistical tests to help analysts decide if the distribution can be considered as normal or multinormal. If the distribution needs to be normalized, an included widget can perform Box-Cox transformations. I also created widgets to detect outliers in the univariate or multivariate distribution. There is a simple widget to test homogeneity of variance (Homoscedasticity) and another to compute confidence intervals. Finally, you will find a widget to create sequences of numbers.

The "compositions" package is made for people dealing with data that are part of a whole, i.e. compositional data. These data can be concentrations of chemicals, human activities in a given territory, probabilities, hours spent in a day for a series of tasks, nutrient needs, etc. Parts of a composition do not lie in the real space: they are confined to a closed space between zero and the unit or the scale of measurement (1, 100%, 1 000 000 ppm, etc.). These data should be analyzed with care since they are subject to numerous inherent limitations, like non-normal distributions, redundancy of information, spurious correlations and scale-dependency. For more information, the website compositionaldata.com can be a good starting point. I created widgets to perform several operations over compositional data. I will not present them in detail here, but feel free to email me if you need further information (my email address is included in the package).

Do not hesitate to contact me for issues, comments or implementations you would like to see.

Regards,

S.-É. Parent
Université Laval
Canada