better testing of family for GLMM conversion (#66)

* more robust familial testing

* extra link cleaning

* patch bump

* remove unused type param

* simplify

* Update src/merMod.jl

Project.toml

@@ -1,7 +1,7 @@
 name = "JellyMe4"
 uuid = "19ac8677-9a15-4623-9afd-84acc6165ce7"
 authors = ["Phillip Alday <[email protected]>"]
-version = "1.1.0"
+version = "1.1.1"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"

src/glmerMod.jl

@@ -151,18 +151,18 @@ function sexp(::Type{RClass{:glmerMod}},
     # we construct this piece-by-piece because this keeps the warnings
     # and everything in sync
     supported = ("Bernoulli", "Binomial", "Poisson")
-    if family in ("Bernoulli", "Binomial")
+    if distribution <: Union{Bernoulli,Binomial}
         family = "binomial"
-    elseif family == "Poisson"
+    elseif distribution <: Poisson
         family = "poisson"
-    elseif family in ("Gamma", "Gaussian", "InverseGaussian")
+    elseif distribution <: Union{Gamma,Gaussian,InverseGaussian}
         throw(ArgumentError("GLMMs with dispersion parameters are known to give incorrect results in MixedModels.jl (see PR#291), aborting."))
     else
         throw(ArgumentError("Family $family is not supported"))
     end
 
     urlink = string(Link(m.resp))
-    link = lowercase(replace(urlink, "Link()" => ""))
+    link = lowercase(replace(replace(urlink, "Link()" => ""), "GLM." => ""))
 
     link in ["logit", "probit", "cauchit",
              "log", "identity", "inverse", "sqrt",
@@ -184,7 +184,7 @@ function sexp(::Type{RClass{:glmerMod}},
         # should we check for PooledArray? brings in another dependency...
         if !(jellyme4_data[!, lhs.sym] isa CategoricalArray)
             @warn "Response was not categorical, converting in place"
-            categorical!(jellyme4_data, [lhs.sym])
+            categorical!(jellyme4_data, lhs.sym)
         end
     end
     formula = convert_julia_to_r(m.formula)

src/merMod.jl

@@ -25,4 +25,4 @@ function sexp(::Type{RClass{:merMod}}, x::Tuple{Union{DataFrame,ColumnTable},Mix
     throw(ArgumentError("The order in your Tuple is reversed. It should be (Model, Data)"))
 end
 
-sexpclass(x::Tuple{Union{DataFrame,ColumnTable},MixedModel}) where {T} = RClass{:merMod}
+sexpclass(x::Tuple{Union{DataFrame,ColumnTable},MixedModel}) = RClass{:merMod}