Enforce type consistency (#43)

* enforce type consistency

src/robustlinearmodel.jl

@@ -98,8 +98,8 @@ function StatsAPI.fit(
     end
 
     # Make sure X and y have the same float eltype
-    T = promote_type(float(eltype(X)), float(eltype(y)))
-    return fit(M, convert.(T, X), convert.(T, y), args...; kwargs...)
+    pX, py = promote_to_same_float(X, y)
+    return fit(M, pX, py, args...; kwargs...)
 end
 
 ## Convert from formula-data to modelmatrix-response calling form
@@ -119,7 +119,10 @@ function StatsAPI.fit(
     # Extract arrays from data using formula
     f, y, X, extra = modelframe(f, data, contrasts, dropmissing, M; wts=wts)
     # Call the `fit` method with arrays
-    return fit(M, X, y, args...; wts=extra.wts, contrasts=contrasts, __formula=f, kwargs...)
+    pX, py = promote_to_same_float(X, y)
+    return fit(
+        M, pX, py, args...; wts=extra.wts, contrasts=contrasts, __formula=f, kwargs...
+    )
 end
 
 
@@ -1021,11 +1024,11 @@ function pirls!(
     devold = deviance(m)
     absdev = abs(devold)
     dev = devold
-    Δdev = 0
+    Δdev = zero(T)
 
     verbose && println("initial deviance: $(@sprintf("%.4g", devold))")
     for i in 1:maxiter
-        f = 1.0 # line search factor
+        f = one(T) # line search factor
         # local dev
         absdev = abs(devold)
 
@@ -1124,12 +1127,12 @@ function pirls_Sestimate!(
     sigold = scale(
         setη!(m; updatescale=true, verbose=verbose, sigma0=sigma0, fallback=maxσ)
     )
-    installbeta!(p, 1)
+    installbeta!(p, one(T))
     r.σ = sigold
 
     verbose && println("initial iteration scale: $(@sprintf("%.4g", sigold))")
     for i in 1:maxiter
-        f = 1.0 # line search factor
+        f = one(T) # line search factor
         local sig
 
         # Compute the change to β, update μ and compute deviance
@@ -1242,11 +1245,11 @@ function pirls_τestimate!(
 
     # Compute initial τ-scale
     tauold = tauscale(setη!(m; updatescale=true); verbose=verbose)
-    installbeta!(p, 1)
+    installbeta!(p, one(T))
 
     verbose && println("initial iteration τ-scale: $(@sprintf("%.4g", tauold))")
     for i in 1:maxiter
-        f = 1.0 # line search factor
+        f = one(T) # line search factor
         local tau
 
         # Compute the change to β, update μ and compute deviance
@@ -1366,6 +1369,7 @@ function resampling_best_estimate(
     ## Hubert2015 - The DetS and DetMM estimators for multivariate location and scatter
     ## (https://www.sciencedirect.com/science/article/abs/pii/S0167947314002175)
     M = length(coef(m))
+    T = eltype(coef(m))
 
     if isnothing(Nsamples)
         Nsamples = resampling_minN(M, 0.05, propoutliers)
@@ -1377,8 +1381,8 @@ function resampling_best_estimate(
 
 
     verbose && println("Start $(Nsamples) subsamples...")
-    σis = zeros(Nsamples)
-    βis = zeros(M, Nsamples)
+    σis = zeros(T, Nsamples)
+    βis = zeros(T, M, Nsamples)
     for i in 1:Nsamples
         # TODO: to parallelize, make a deepcopy of m
         inds = sample(rng, axes(response(m), 1), Npoints; replace=false, ordered=false)
@@ -1393,7 +1397,7 @@ function resampling_best_estimate(
         # Initialize σ as mad(residuals)
         setinitσ!(m)
 
-        σi = 0
+        σi = zero(T)
         for k in 1:Nsteps_β
             setη!(
                 m;

src/tools.jl

@@ -4,6 +4,17 @@
 ##    Missing values
 ################################################
 
+function promote_to_same_float(X::AbstractMatrix, y::AbstractVector)
+    T = promote_type(float(eltype(X)), float(eltype(y)))
+    if !(T <: AbstractFloat)
+        msg = "promoting X and y arrays to float types"
+        throw(TypeError(:fit, msg, Type{<:AbstractFloat}, T))
+    end
+    MT = AbstractMatrix{T}
+    VT = AbstractVector{T}
+    return convert.(T, X)::MT, convert.(T, y)::VT
+end
+
 _missing_omit(x::AbstractArray{T}) where {T} = copyto!(similar(x, nonmissingtype(T)), x)
 
 function StatsModels.missing_omit(X::AbstractMatrix, y::AbstractVector)