agasitko

Cluster coloring with non-equidistant nodes

2

There is a bug in the coloring (`plot.factorMerger` with the argument `statistic = "p-value"`). See below:  Needs to be fixed.

Optimizition: use lm.fit instead of lm

2

lm.fit is faster than lm. However, some statistics must be calculated separately (e.g. group means).

Make right hand side of the tree plot more visible

2

The most interesting part of the merging path is usually on the right hand side of the tree plot. Loglikelihood varies only slightly for early models and it's hard to...

Align plots

1

`plot.factorMerger` should have panels align by: - x axis (GIC and tree), - y axis (tree and response panel - if possible)

cv.mergeFactors()

1

Why not to add a function that will run CV to choose the optimal penalty in GIC for a given dataset?