economistgame / rcmdrplugin.optimclassifier Goto Github PK

When running Optim.LMM() after loading RcmdrPlugin.OptimClassifier, it occasionally gives errors. I'm guessing there's some sort of package namespace conflict behind the scenes, but I can't uncover it. The model works fine when run manually with just the OptimClassifier library.

Here's a reproducible example using OptimClassifier 0.1.4 and RcmdrPlugin.OptimClassifier 0.1.2:

# This works
library(OptimClassifier)
modelFit <- Optim.LMM("Y", AustralianCredit, p = 0.7, seed = 2018)
summary(modelFit)

# Start a new session of R. This doesn't work
library(RcmdrPlugin.OptimClassifier)
modelFit <- Optim.LMM("Y", AustralianCredit, p = 0.7, seed = 2018)
#> Error in eval(predvars, data, env) : object 'X4' not found

Review issue: openjournals/joss-reviews#727

When running "Models" > "Graphs" > "Basic diagnostic plots" on a discriminant analysis model (or when just running plot(), plotting fails.

OptimDA2 <- Optim.DA(Y~X1+X2+X3+X4+X5+X6+X7+X8+X9+X10+X11+X12+X13+X14, 
                     data=AustralianCredit, p=0.7, criteria='rmse',seed=2018)
#> Error in rep("darkgrey", nrow(Files) - 2) : invalid 'times' argument

Review issue: openjournals/joss-reviews#727

This may be an issue with OptimClassifier 0.1.4, but it also has a Rcmdr component. OptimClassifier::Optim.LM() says that it takes a formula as an argument, but when you pass a formula, it complains that the formula object is not a matrix:

Optim.LM(Y ~ X1 + X2 + X3 + X4 + X5 + X6, data = AustralianCredit, p = 0.7, 
         criteria = 'success_rate', seed = 2018)
#> Error in model.frame.default(formula = ModelsTested$train_formula[[i]],  : 
#> object is not a matrix

As a result, it's not currently possible to select a few variables using R Commander. For instance, this:

…results in this:

Optim.LM() does work if the formula uses a . for the RHS, but there's no way to generate that with R Commander—selecting all the variables just concatenates them into a long formula:

# This works, but can't be generated with R Commander
Optim.LM(Y~., AustralianCredit, p = 0.7, seed = 2018)

# This doesn't work, but it's what R Commander generates
# Optim.LM(Y~X1+X2+X3+X4+X5+X6+X7+X8+X9+X10+X11+X12+X13+X14, 
#          AustralianCredit, p = 0.7, seed = 2018)

Review issue: openjournals/joss-reviews#727

There's currently an empty/corrupt Rplots.pdf in the root of the repository that doesn't need to be there and can be removed from the repo.

Review issue: openjournals/joss-reviews#727