abstract near-zero checks into configurable params

NEWS.md

@@ -6,6 +6,7 @@ MixedModels v4.24.2 Release Notes
     - user-specified `σ` is actually used, defaulting to existing value
     - `REML` defaults to model's already specified REML value.
 - Clean up code of keyword convenience constructor for `OptSummary`. [#772]
+- Refactor thresholding parameters for forcing near-zero parameter values into `OptSummary`. [#772]
 
 MixedModels v4.24.1 Release Notes
 ==============================

src/generalizedlinearmixedmodel.jl

@@ -309,13 +309,13 @@ function StatsAPI.fit!(
     ## check if very small parameter values bounded below by zero can be set to zero
     xmin_ = copy(xmin)
     for i in eachindex(xmin_)
-        if iszero(optsum.lowerbd[i]) && zero(T) < xmin_[i] < T(0.001)
+        if iszero(optsum.lowerbd[i]) && zero(T) < xmin_[i] < optsum.xtol_zero_abs
             xmin_[i] = zero(T)
         end
     end
     loglength = length(fitlog)
     if xmin ≠ xmin_
-        if (zeroobj = obj(xmin_, T[])) ≤ (fmin + 1.e-5)
+        if (zeroobj = obj(xmin_, T[])) ≤ (fmin + optsum.ftol_zero_abs)
             fmin = zeroobj
             copyto!(xmin, xmin_)
         elseif length(fitlog) > loglength

src/linearmixedmodel.jl

@@ -512,13 +512,13 @@ function StatsAPI.fit!(
     xmin_ = copy(xmin)
     lb = optsum.lowerbd
     for i in eachindex(xmin_)
-        if iszero(lb[i]) && zero(T) < xmin_[i] < T(0.001)
+        if iszero(lb[i]) && zero(T) < xmin_[i] < optsum.xtol_zero_abs
             xmin_[i] = zero(T)
         end
     end
     loglength = length(fitlog)
     if xmin ≠ xmin_
-        if (zeroobj = obj(xmin_, T[])) ≤ (fmin + 1.e-5)
+        if (zeroobj = obj(xmin_, T[])) ≤ (fmin + optsum.ftol_zero_abs)
             fmin = zeroobj
             copyto!(xmin, xmin_)
         elseif length(fitlog) > loglength

src/optsummary.jl

@@ -19,6 +19,7 @@ Summary of an `NLopt` optimization
 * `feval`: the number of function evaluations
 * `optimizer`: the name of the optimizer used, as a `Symbol`
 * `returnvalue`: the return value, as a `Symbol`
+* `xtol_zero_abs`: the tolerance for a near zero parameter to be considered practically zero
 * `ftol_zero_abs`: the tolerance for change in the objective for setting a near zero parameter to zero
 * `fitlog`: A vector of tuples of parameter and objectives values from steps in the optimization
 * `nAGQ`: number of adaptive Gauss-Hermite quadrature points in deviance evaluation for GLMMs
@@ -50,7 +51,8 @@ Base.@kwdef mutable struct OptSummary{T<:AbstractFloat}
     fmin::T = Inf * one(eltype(initial))
     optimizer::Symbol = :LN_BOBYQA
     returnvalue::Symbol = :FAILURE
-    ftol_zero_abs::T = one(eltype(initial)) / 1000
+    xtol_zero_abs::T = eltype(initial)(0.001)
+    ftol_zero_abs::T = eltype(initial)(1.e-5)
     # not SVector because we would need to parameterize on size (which breaks GLMM)
     fitlog::Vector{Tuple{Vector{T},T}} = [(initial, fmin)]
     # don't really belong here but I needed a place to store them