The Nash-Sutcliffe efficiency , Willmott's index and Model Performance Index (MPI) are suitable indices to add XLSTAT for assessing model performance.

When performing the Kruskal-Wallis analysis of variance on k samples, up until now, you can choose between Dunn, Conover-Iman and Steel-Dwass-Critchlow-Fligner with the addition of Bonferroni correction (for the first two mentioned, respectively) ...

Could you please add XGboost as a new Machine Learning available algorithm?

I think XLSTAT provides partial least squares models but not general least squares. This would be very practical for time series analysis, for instance.

Rolling multiple regression is extremely useful in that it enables you to see how stable the beta coefficients are over time. This is a very common issue that comes up regularly when you deal with financial time series.

Spatial Autocorrelation Indices (e.g Moran's I) are mainstream of spatial statistics that are worth being implemented in XLSTAT software. They are useful for ecological studies.

Would be nice to be able to integrate the area under the curve in linear and non-linear models.

Estimating lagged prediction models from autocorrelated historical data is a common challenge in business and economics. The resulting multicollinearity causes the models to perform poorly. The Almon lag method is a clever way to deal with the mul...

Here is a series of references for checking for groups of outliers in the data: https://lemakistatheux.wordpress.com/category/tests-statistique-indices-de-liaison-et-coefficients-de-correlation/les-tests-pour-la-detection-doutliers/

The scatterplot matrix is a great way to quickly visualize correlations among multivariate data. It is especially helpful if confidence ellipses can be added and if the axes can be easily edited.