Stop relying on TableRegressionModel

[nalimilan]

We're going to stop using TableRegressionModel in GLM.jl (JuliaStats/GLM.jl#339) as it's really not convenient for users to have that wrapper instead of just working with LinearModel and GeneralizedLinearModel. TableRegressionModel will probably be deprecated soon in StatsModels.

I see that RegressionTables.jl is one of the few packages that use TableRegressionModel. In particular standardized coefficients are only supported for these models:

RegressionTables.jl/src/regtable.jl

Lines 197 to 198 in e13df11

	if standardize_coef && isa(rr[resultIndex],StatsModels.TableRegressionModel)
	thiscoef = [ thiscoef[i]*std(rr[resultIndex].model.pp.X[:,i])/std(rr[resultIndex].model.rr.y) for i in 1:length(thiscoef) ]

IIUC, you use this to detect that some models are GLM.jl objects and access their private fields. I suspect you could achieve the same result for any model (not just GLM) by calling modelmatrix and response instead. Can you confirm this? It would be too bad to lose support for GLM.jl in RegressionTables.jl.

[jmboehm]

Yes, I've been meaning for a while to write this in a more general way. In any case I'll make sure that GLM.jl will always be supported.