get_probabilities: Probabilities for multinomial regression
Description
Function which computes probabilities of outcomes on the test set by applying regression parameters inferred by a run on the training set. Works for logistic or multinomial regression
Probabilities of the outcomes. In 'binary' mode returns an array of the size of the number of observations in a testset. In 'multin' returns an M x N matrix where M is the size of the number of observations in a testset
and N is the number of unique outcomes minus 1.
Arguments
trset
values of predictors on the training set
testset
values of predictors on the test set
outc
values of outcomes on the training set
mode
'binary' (logistic regression) or 'multin' (multinomial regression)
Rsq
whether R-squared statistics constrained is introduced
p
weight of the model
n_tr
size of the training set
Details
In binary mode this function computes the probabilities of the event '0'. In multinomial mode computes the probabilities of the events '0','1',...,'N-1'.