Update for logistic, negative binomial regression

[stueller]

• The function cutY in utils.r is a wrapper for the cut function that cleans the output into a numeric variable.
  • In the code for polyICT there is an error for cutting y into more than five categories because 5 is the maximum supported by multinomial regression.
  • This could be changed to a warning and the 'breaks' (which must sum to 1) could be specified to match a distribution of counts obtained from rNBI or rnbinom
  • There is an error in the cutY code as.numeric(factor(u, labels = 1:length(yCut))) - 1, the categorization is correct but the values are non-integer (the code looks correct, check for downstream bugs).
• A polyICT object can generate categorical data (up to 5 categories for now) using $makeData(), but $designCheck() has an error when yCut is not NULL.
• PersonAlytic fails with the error "x must be between 0 and 1", something must be going wrong in the call to the gamlss function fitDist in Palytic.r, most likely the option real0to1 is being included when it should not be. This would happen when the parameter is01 on line 1447 in the PersonAlytics Palytic$set() definition in the Palytic.r file. Set autoSelect = list() in PersonAlytic() as a temporary workaround.
• ICTpower() generates data with a polyICT object where yCut!=NULL but yCut is ignored. Passing a family parameter throws the error formal argument "family" matched by multiple actual arguments.

[stueller]

• mean centering of the outcome is default for makeData override with $makeData(y0atyMean=F)
• the fitDist issue is fixed changing is01 <- min(dv, na.rm=TRUE) >= 0 & max(dv, na.rm=TRUE) <= 1 to is01 <- min(dv, na.rm=TRUE) >= 0 & max(dv, na.rm=TRUE) <= 1 & !isBin