finish arg checks

Author: itsdfish

src/alternative_geometries/CircularDDM.jl

@@ -46,6 +46,14 @@ struct CDDM{T <: Real} <: AbstractCDDM
     η::Vector{T}
     α::T
     τ::T
+
+    function CDDM(ν::Vector{T}, σ::T, η::Vector{T}, α::T, τ::T) where {T}
+        @argcheck σ ≥ 0
+        @argcheck all(η .≥ 0)
+        @argcheck α ≥ 0
+        @argcheck τ ≥ 0
+        return new{T}(ν, σ, η, α, τ)
+    end
 end
 
 function CDDM(ν, σ, η, α, τ)
@@ -55,9 +63,7 @@ function CDDM(ν, σ, η, α, τ)
     return CDDM(ν, σ, η, α, τ)
 end
 
-function params(d::AbstractCDDM)
-    return (d.ν, d.σ, d.η, d.α, d.τ)
-end
+params(d::AbstractCDDM) = (d.ν, d.σ, d.η, d.α, d.τ)
 
 function CDDM(; ν = [1, 0.5], η = [1, 1], σ = 1, α = 1.5, τ = 0.30)
     return CDDM(ν, σ, η, α, τ)
@@ -94,6 +100,7 @@ end
 
 function logpdf(d::AbstractCDDM, data::Vector{<:Real}; k_max = 50)
     θ, rt = data
+    @argcheck d.τ ≤ rt
     return logpdf_term1(d, θ, rt) + logpdf_term2(d, rt; k_max)
 end
 
@@ -111,11 +118,13 @@ end
 
 function pdf(d::AbstractCDDM, data::Vector{<:Real}; k_max = 50)
     θ, rt = data
+    @argcheck d.τ ≤ rt
     return max(0.0, pdf_term1(d, θ, rt) * pdf_term2(d, rt; k_max))
 end
 
 function pdf(d::AbstractCDDM, data::Vector{<:Real}, j0, j01, j02; k_max = 50)
     θ, rt = data
+    @argcheck d.τ ≤ rt
     return max(0.0, pdf_term1(d, θ, rt) * pdf_term2(d, rt, j0, j01, j02; k_max))
 end
 

src/multi_choice_models/ClassicMDFT.jl

@@ -6,18 +6,25 @@ A model type for Multiattribute Decision Field Theory.
     
 # Parameters 
 - `σ::T = 1.0`: diffusion noise. σ ∈ ℝ⁺. 
-- `α::T = 15.0`: evidence threshold. α  ∈ ℝ⁺.
-- `τ::T = .30`: non-decision time. τ ∈ [0, min_rt].
-- `w::Vector{T}`: attention weights vector where each element corresponds to the attention given to the corresponding dimension. wᵢ ∈ [0,1], ∑wᵢ = 1.
+- `C::Array{T, 2}`: contrast weight matrix where c_ij is the contrast weight when comparing options i and j.
 - `S::Array{T, 2}`: feedback matrix allowing self-connections and interconnections between alternatives. Self-connections range from zero to 1, where s_ij < 1 represents decay. Interconnections 
      between options i and j  where i ≠ j are inhibatory if s_ij < 0.
-- `C::Array{T, 2}`: contrast weight matrix where c_ij is the contrast weight when comparing options i and j.
+- `w::Vector{T}`: attention weights vector where each element corresponds to the attention given to the corresponding dimension. wᵢ ∈ [0,1], ∑wᵢ = 1.
+- `α::T = 15.0`: evidence threshold. α  ∈ ℝ⁺.
+- `τ::T = .30`: non-decision time. τ ∈ [0, min_rt].
 
 # Constructors 
 
-    ClassicMDFT(σ, α, τ, w, S, C)
+    ClassicMDFT(σ, C, S, w, α, τ)
 
-    ClassicMDFT(σ, α, τ, w, S, C = make_default_contrast(S))
+    ClassicMDFT(;
+        σ = 1.0,
+        α = 15.0,
+        τ = 10.0,
+        w,
+        S,
+        C = make_default_contrast(size(S, 1))
+    )
         
 # Example 
 
@@ -34,12 +41,8 @@ M = [
 ]
 
 model = ClassicMDFT(;
-    # non-decision time 
-    τ = 0.300,
     # diffusion noise 
     σ = 1.0,
-    # decision threshold
-    α = 17.5,
     # attribute attention weights 
     w = [0.5, 0.5],
     # feedback matrix 
@@ -48,6 +51,10 @@ model = ClassicMDFT(;
         -0.0122316 0.9500000 -0.00903030
         -0.0499996 -0.0090303 0.95000000
     ],
+    # decision threshold
+    α = 17.5,
+    # non-decision time 
+    τ = 0.300,
 )
 choices, rts = rand(model, 10_000, M; Δt = 1.0)
 map(c -> mean(choices .== c), 1:3)
@@ -58,19 +65,33 @@ Roe, Robert M., Jermone R. Busemeyer, and James T. Townsend. "Multiattribute Dec
 """
 mutable struct ClassicMDFT{T <: Real} <: AbstractMDFT
     σ::T
+    C::Array{T, 2}
+    S::Array{T, 2}
+    w::Vector{T}
     α::T
     τ::T
-    w::Vector{T}
-    S::Array{T, 2}
-    C::Array{T, 2}
+    function ClassicMDFT(
+        σ::T,
+        C::Array{T, 2},
+        S::Array{T, 2},
+        w::Vector{T},
+        α::T,
+        τ::T
+    ) where {T <: Real}
+        @argcheck σ ≥ 0
+        @argcheck α ≥ 0
+        @argcheck τ ≥ 0
+        @argcheck all(w .≥ 0) && (sum(w) == 1)
+        return new{T}(σ, C, S, w, α, τ)
+    end
 end
 
-function ClassicMDFT(σ, α, τ, w, S, C)
-    σ, α, τ, _, _, _ = promote(σ, α, τ, w[1], S[1], C[1])
-    w = convert(Vector{typeof(τ)}, w)
-    S = convert(Array{typeof(τ), 2}, S)
+function ClassicMDFT(σ, C, S, w, α, τ)
+    σ, _, _, _, α, τ, = promote(σ, C[1], S[1], w[1], α, τ)
     C = convert(Array{typeof(τ), 2}, C)
-    return ClassicMDFT(σ, α, τ, w, S, C)
+    S = convert(Array{typeof(τ), 2}, S)
+    w = convert(Vector{typeof(τ)}, w)
+    return ClassicMDFT(σ, C, S, w, α, τ)
 end
 
 function ClassicMDFT(;
@@ -81,14 +102,12 @@ function ClassicMDFT(;
     S,
     C = make_default_contrast(size(S, 1))
 )
-    return ClassicMDFT(σ, α, τ, w, S, C)
+    return ClassicMDFT(σ, C, S, w, α, τ)
 end
 
-get_pdf_type(d::AbstractMDFT) = Approximate
+get_pdf_type(::AbstractMDFT) = Approximate
 
-function params(d::ClassicMDFT)
-    return (d.σ, d.α, d.τ, d.w, d.S, d.C)
-end
+params(d::ClassicMDFT) = (d.σ, d.C, d.S, d.w, d.α, d.τ)
 
 """
     rand(

src/multi_choice_models/DDM.jl

@@ -39,15 +39,19 @@ mutable struct DDM{T <: Real} <: AbstractDDM
     α::T
     z::T
     τ::T
+    function DDM(ν::T, α::T, z::T, τ::T) where {T <: Real}
+        @argcheck α ≥ 0
+        @argcheck (z ≥ 0) && (z ≤ 1)
+        @argcheck τ ≥ 0
+        return new{T}(ν, α, z, τ)
+    end
 end
 
 function DDM(ν, α, z, τ)
     return DDM(promote(ν, α, z, τ)...)
 end
 
-function params(d::DDM)
-    return (d.ν, d.α, d.z, d.τ)
-end
+params(d::DDM) = (d.ν, d.α, d.z, d.τ)
 
 function DDM(; ν = 1.00, α = 0.80, τ = 0.30, z = 0.50)
     return DDM(ν, α, z, τ)
@@ -64,7 +68,8 @@ end
 # Wabersich & Vandekerckhove (2014) #
 #####################################
 
-function pdf(d::DDM, choice, rt; ϵ::Real = 1.0e-12)
+function pdf(d::DDM{T}, choice, rt; ϵ::Real = 1.0e-12) where {T <: Real}
+    @argcheck d.τ < rt
     if choice == 1
         (ν, α, z, τ) = params(d)
         return _pdf(DDM(-ν, α, 1 - z, τ), rt; ϵ)
@@ -73,11 +78,8 @@ function pdf(d::DDM, choice, rt; ϵ::Real = 1.0e-12)
 end
 
 # probability density function over the lower boundary
-function _pdf(d::DDM{T}, t::Real; ϵ::Real = 1.0e-12) where {T <: Real}
+function _pdf(d::DDM, t::Real; ϵ::Real = 1.0e-12)
     (ν, α, z, τ) = params(d)
-    if τ ≥ t
-        return T(NaN)
-    end
     u = (t - τ) / α^2 #use normalized time
 
     K_s = 2.0
@@ -97,7 +99,6 @@ function _pdf(d::DDM{T}, t::Real; ϵ::Real = 1.0e-12) where {T <: Real}
     if K_s < K_l
         return p * _small_time_pdf(u, z, ceil(Int, K_s))
     end
-
     return p * _large_time_pdf(u, z, ceil(Int, K_l))
 end
 
@@ -106,7 +107,7 @@ function _small_time_pdf(u::T, z::T, K::Int) where {T <: Real}
     inf_sum = zero(T)
 
     k_series = (-floor(Int, 0.5 * (K - 1))):ceil(Int, 0.5 * (K - 1))
-    for k in k_series
+    for k ∈ k_series
         inf_sum += ((2k + z) * exp(-((2k + z)^2 / (2u))))
     end
 
@@ -117,15 +118,14 @@ end
 function _large_time_pdf(u::T, z::T, K::Int) where {T <: Real}
     inf_sum = zero(T)
 
-    for k = 1:K
+    for k ∈ 1:K
         inf_sum += (k * exp(-0.5 * (k^2 * π^2 * u)) * sin(k * π * z))
     end
 
     return π * inf_sum
 end
 
 logpdf(d::DDM, choice, rt; ϵ::Real = 1.0e-12) = log(pdf(d, choice, rt; ϵ))
-#logpdf(d::DDM, t::Real; ϵ::Real = 1.0e-12) = log(pdf(d, t; ϵ))
 
 logpdf(d::DDM, data::Tuple) = logpdf(d, data...)
 
@@ -134,7 +134,8 @@ logpdf(d::DDM, data::Tuple) = logpdf(d, data...)
 # Blurton, Kesselmeier, & Gondan (2012) #
 #########################################
 
-function cdf(d::DDM, choice::Int, rt::Real = 10; ϵ::Real = 1.0e-12)
+function cdf(d::DDM{T}, choice::Int, rt::Real = 10; ϵ::Real = 1.0e-12) where {T <: Real}
+    @argcheck d.τ < rt
     if choice == 1
         (ν, α, z, τ) = params(d)
         return _cdf(DDM(-ν, α, 1 - z, τ), rt; ϵ)
@@ -144,11 +145,7 @@ function cdf(d::DDM, choice::Int, rt::Real = 10; ϵ::Real = 1.0e-12)
 end
 
 # cumulative density function over the lower boundary
-function _cdf(d::DDM{T}, t::Real; ϵ::Real = 1.0e-12) where {T <: Real}
-    if d.τ ≥ t
-        return T(NaN)
-    end
-
+function _cdf(d::DDM, t::Real; ϵ::Real = 1.0e-12)
     K_l = _K_large(d, t; ϵ)
     K_s = _K_small(d, t; ϵ)
 
@@ -163,7 +160,7 @@ function _Fl_lower(d::DDM{T}, K::Int, t::Real) where {T <: Real}
     (ν, α, z, τ) = params(d)
     F = zero(T)
     K_series = K:-1:1
-    for k in K_series
+    for k ∈ K_series
         F -= (
             k / (ν^2 + k^2 * π^2 / (α^2)) *
             exp(-ν * α * z - 0.5 * ν^2 * (t - τ) - 0.5 * k^2 * π^2 / (α^2) * (t - τ)) *
@@ -186,7 +183,7 @@ function _Fs_lower(d::DDM{T}, K::Int, t::Real) where {T <: Real}
     S2 = zero(T)
     K_series = K:-1:1
 
-    for k in K_series
+    for k ∈ K_series
         S1 += (
             _exp_pnorm(2 * ν * α * k, -sign(ν) * (2 * α * k + α * z + ν * (t - τ)) / sqt) - _exp_pnorm(
             -2 * ν * α * k - 2 * ν * α * z,
@@ -214,11 +211,10 @@ function _Fs_lower(d::DDM{T}, K::Int, t::Real) where {T <: Real}
 end
 
 # Zero drift version
-function _Fs0_lower(d::DDM{T}, K::Int, t::Real) where {T <: Real}
-    (_, α, z, τ) = params(d)
+function _Fs0_lower(dist::DDM{T}, K::Int, t::Real) where {T <: Real}
+    (; α, z, τ) = dist
     F = zero(T)
-    K_series = K:-1:0
-    for k in K_series
+    for k ∈ K:-1:0
         F -= (
             cdf(Distributions.Normal(), (-2 * k - 2 + z) * α / sqrt(t - τ)) +
             cdf(Distributions.Normal(), (-2 * k - z) * α / sqrt(t - τ))

src/multi_choice_models/LBA.jl

@@ -41,12 +41,18 @@ mutable struct LBA{T <: Real, T1 <: Union{<:T, Vector{<:T}}} <: AbstractLBA{T, T
     A::T
     k::T
     τ::T
-    function LBA(ν::Vector{T}, σ::T1, A::T, k::T, τ::T) where {T <: Real, T1 <: Union{<:T, Vector{<:T}}}
+    function LBA(
+        ν::Vector{T},
+        σ::T1,
+        A::T,
+        k::T,
+        τ::T
+    ) where {T <: Real, T1 <: Union{<:T, Vector{<:T}}}
         @argcheck all(σ .≥ 0)
         @argcheck A ≥ 0
         @argcheck k ≥ 0
         @argcheck τ ≥ 0
-        return new{T,T1}(ν, σ, A, k, τ)
+        return new{T, T1}(ν, σ, A, k, τ)
     end
 end
 
@@ -84,7 +90,6 @@ sample_drift_rates(ν, σ) = sample_drift_rates(Random.default_rng(), ν, σ)
 function sample_drift_rates(rng::AbstractRNG, ν, σ)
     negative = true
     v = similar(ν)
-    n_options = length(ν)
     while negative
         v = @. rand(rng, Normal(ν, σ))
         negative = any(x -> x > 0, v) ? false : true
@@ -106,8 +111,8 @@ end
 
 function logpdf(d::AbstractLBA{T, T1}, c, rt) where {T, T1 <: Vector{<:Real}}
     (; τ, ν, σ) = d
+    @argcheck τ ≤ rt
     LL = 0.0
-    rt < τ ? (return -Inf) : nothing
     for i ∈ 1:length(ν)
         if c == i
             LL += log_dens(d, ν[i], σ[i], rt)
@@ -123,7 +128,7 @@ end
 function logpdf(d::AbstractLBA{T, T1}, c, rt) where {T, T1 <: Real}
     (; τ, ν, σ) = d
     LL = 0.0
-    rt < τ ? (return -Inf) : nothing
+    @argcheck τ ≤ rt
     for i ∈ 1:length(ν)
         if c == i
             LL += log_dens(d, ν[i], σ, rt)
@@ -137,9 +142,9 @@ function logpdf(d::AbstractLBA{T, T1}, c, rt) where {T, T1 <: Real}
 end
 
 function pdf(d::AbstractLBA{T, T1}, c, rt) where {T, T1 <: Vector{<:Real}}
-    (; τ, A, k, ν, σ) = d
+    (; τ, ν, σ) = d
     den = 1.0
-    rt < τ ? (return 1e-10) : nothing
+    @argcheck τ ≤ rt
     for i ∈ 1:length(ν)
         if c == i
             den *= dens(d, ν[i], σ[i], rt)
@@ -156,7 +161,7 @@ end
 function pdf(d::AbstractLBA{T, T1}, c, rt) where {T, T1 <: Real}
     (; τ, ν, σ) = d
     den = 1.0
-    rt < τ ? (return 1e-10) : nothing
+    @argcheck τ ≤ rt
     for i ∈ 1:length(ν)
         if c == i
             den *= dens(d, ν[i], σ, rt)