a gateway to R constructs would open up the capability of xlstat

While PCA is used to explore the relation between several quantitative variables, MCA (Multiple Correspondance Analysis) is used to explore the relation between several categorical variables. Some methods (PCAmix method for exemple) can deal with ...

In the {Significance by cell (Fisher's exact test)} table, we can only get < or >.It would be valuable to have the corresponding p-values per cell.

Random Forests is a mainstream machine learning algorithm that is worth being implemented in XLSTAT. Maybe the XLSTAT decision trees code could be used as a baseline to develop Random Forests?

The quality of an ANCOVA depends on the heteroscedasticity and the normality of residuals. Currently, these two aspects are not part of the output of the ANCOVA analysis. They have to be manually calculated (in the basic version for normality, in ...

For nonlinear regressions add Akaike Information Criteria for finite samples goodness of fit

Hello, it would be interesting if you add normality test and homogeneous test before ANOVA test. In this case, XLSTAT could say "You can use ANOVA because your data respect hypothesis of ANOVA" or "Your data are not adapted for ANOVA test, do you ...

The XLSTAT non parametric tests list is quite complete. One essential feature is missing, though: the Wilcoxon one-sample test to compare the location of a sample to a hypothesized value.

Computing Monte-Carlo simulated p-values for t-tests or ANOVA is an interesting non-parametric approach to perform these tests using the real "raw" data instead of ranks. This approach is implemented in the coin package in R.

The attached image speaks for itself: each pair of bars is a dependent variable in the context of a one-way ANOVA model (factor: sample with two levels). The factor has a significant effect on the output for Attribute 1.