Glm additions

[rsenne]

This pull request significantly expands and improves the documentation for EmissionModels.jl, adds new emission models and priors, and introduces several usability and performance enhancements to the codebase. The main changes include a comprehensive rewrite of the README.md, the addition of detailed documentation pages for distributions, GLM emissions, priors, and custom model interfaces, as well as new exports and improved internal handling for multivariate t-distributions.

Documentation improvements:

• Major rewrite and expansion of README.md to include a clear package overview, quick start, model summaries, installation, and contribution guidelines. [1] [2]
• Added new documentation pages: distributions, GLM emissions, priors, and custom emission models, with detailed usage examples and interface requirements. [1] [2] [3] [4] [5] [6]

Feature additions:

• Exported new emission models and priors, including GaussianGLM, BernoulliGLM, PoissonGLM, multivariate GLMs, and prior types such as RidgePrior, with their associated penalty and gradient functions. [1] [2]
• Added support for GLM emission models and regularization priors in the main module.

Multivariate t-distribution improvements:

• Enhanced MultivariateT and MultivariateTDiag structs to include internal scratch buffers for more efficient and allocation-free fit! and density computations. [1] [2] [3] [4]
• Improved logdensityof implementations for both full and diagonal multivariate t-distributions to reduce allocations and improve thread safety. [1] [2]

General usability:

• Updated documentation and code to clarify and enforce the emission model interface required by HiddenMarkovModels.jl, including sampling, density evaluation, and parameter fitting.

---

References:

• [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

[github-actions]

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

JuliaFormatter

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 565 in 2ae43af

Σ_chol::Cholesky{T, Matrix{T}}

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 572 to 575 in 2ae43af

	B::AbstractMatrix{T},
	Σ::AbstractMatrix{T},
	prior::P,
	) where {T<:Real, P<:AbstractPrior}

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 579 to 580 in 2ae43af

	size(Σ) == (k, k) ||
	throw(DimensionMismatch("Σ must be $(k)×$(k) for B with $(k) columns, got $(size(Σ))"))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 588 to 590 in 2ae43af

	return MvGaussianGLM{T, P}(
	Matrix{T}(B), Matrix{T}(Σ), prior,
	Σ_chol, logdet(Σ_chol), p, k,

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 594 to 595 in 2ae43af

	MvGaussianGLM(B::AbstractMatrix{T}, Σ::AbstractMatrix{T}) where {T<:Real} =
	MvGaussianGLM(B, Σ, NoPrior())

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 618 to 620 in 2ae43af

	glm::MvGaussianGLM,
	y::AbstractVector;
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 624 to 627 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 653 to 655 in 2ae43af

	rng::AbstractRNG,
	glm::MvGaussianGLM{T};
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 680 to 683 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 718 to 719 in 2ae43af

	length(weight_seq) == n ||
	throw(DimensionMismatch("weight_seq length $(length(weight_seq)) ≠ control_seq rows $n"))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 733 to 736 in 2ae43af

	throw(DimensionMismatch(
	"obs_seq[$i] length $(length(obs_i)) ≠ out_dim $k",
	))
	w = T(weight_seq[i])

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 740 in 2ae43af

xa = x_i[a]

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 773 in 2ae43af

w = T(weight_seq[i])

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 823 in 2ae43af

mutable struct MvBernoulliGLM{T<:Real, P<:AbstractPrior} <: AbstractGLM

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 830 in 2ae43af

function MvBernoulliGLM(B::AbstractMatrix{T}, prior::P) where {T<:Real, P<:AbstractPrior}

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 834 in 2ae43af

return MvBernoulliGLM{T, P}(Matrix{T}(B), prior, p, k)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 841 to 842 in 2ae43af

	MvBernoulliGLM(B::AbstractMatrix, prior::AbstractPrior) =
	MvBernoulliGLM(float.(B), prior)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 847 to 849 in 2ae43af

	glm::MvBernoulliGLM,
	y::AbstractVector;
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 853 to 856 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 860 in 2ae43af

for j in 1:glm.out_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 864 in 2ae43af

for r in 1:glm.in_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 873 to 875 in 2ae43af

	rng::AbstractRNG,
	glm::MvBernoulliGLM;
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 896 to 899 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 902 in 2ae43af

for j in 1:glm.out_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 904 in 2ae43af

for r in 1:glm.in_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 932 to 933 in 2ae43af

	length(weight_seq) == n ||
	throw(DimensionMismatch("weight_seq length $(length(weight_seq)) ≠ control_seq rows $n"))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 939 to 942 in 2ae43af

	length(obs_seq[i]) == k ||
	throw(DimensionMismatch(
	"obs_seq[$i] length $(length(obs_seq[i])) ≠ out_dim $k",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 946 to 948 in 2ae43af

	g = Vector{T}(undef, p)
	H = Matrix{T}(undef, p, p)
	Δ = Vector{T}(undef, p)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 954 to 956 in 2ae43af

	β_buf, g, H, Δ, _bernoulli_loss, _bernoulli_gh!,
	yview, weight_seq, control_seq, glm.prior;
	max_iter=max_iter, gtol=gtol, max_backtrack=max_backtrack,

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 978 in 2ae43af

mutable struct MvPoissonGLM{T<:Real, P<:AbstractPrior} <: AbstractGLM

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 985 in 2ae43af

function MvPoissonGLM(B::AbstractMatrix{T}, prior::P) where {T<:Real, P<:AbstractPrior}

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 989 in 2ae43af

return MvPoissonGLM{T, P}(Matrix{T}(B), prior, p, k)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 996 to 997 in 2ae43af

	MvPoissonGLM(B::AbstractMatrix, prior::AbstractPrior) =
	MvPoissonGLM(float.(B), prior)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1002 to 1004 in 2ae43af

	glm::MvPoissonGLM,
	y::AbstractVector;
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1008 to 1011 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 1015 in 2ae43af

for j in 1:glm.out_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 1019 in 2ae43af

for r in 1:glm.in_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1028 to 1030 in 2ae43af

	rng::AbstractRNG,
	glm::MvPoissonGLM;
	control_seq::AbstractVector{<:Real},

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1050 to 1053 in 2ae43af

	length(control_seq) == glm.in_dim ||
	throw(DimensionMismatch(
	"control_seq length $(length(control_seq)) ≠ in_dim $(glm.in_dim)",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 1056 in 2ae43af

for j in 1:glm.out_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Line 1058 in 2ae43af

for r in 1:glm.in_dim

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1086 to 1087 in 2ae43af

	length(weight_seq) == n ||
	throw(DimensionMismatch("weight_seq length $(length(weight_seq)) ≠ control_seq rows $n"))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1093 to 1096 in 2ae43af

	length(obs_seq[i]) == k ||
	throw(DimensionMismatch(
	"obs_seq[$i] length $(length(obs_seq[i])) ≠ out_dim $k",
	))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1100 to 1102 in 2ae43af

	g = Vector{T}(undef, p)
	H = Matrix{T}(undef, p, p)
	Δ = Vector{T}(undef, p)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/glms/glm.jl

Lines 1108 to 1110 in 2ae43af

	β_buf, g, H, Δ, _poisson_loss, _poisson_gh!,
	yview, weight_seq, control_seq, glm.prior;
	max_iter=max_iter, gtol=gtol, max_backtrack=max_backtrack,

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/src/multivariate/t.jl

Line 317 in 2ae43af

for j in 1:d, k in 1:j-1

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Lines 25 to 32 in 2ae43af

	bench_logd(d, y, x, n) = (s = 0.0; for _ in 1:n; s += logdensityof(d, y; control_seq=x); end; s)
	bench_logd_unctrl(d, y, n) = (s = 0.0; for _ in 1:n; s += logdensityof(d, y); end; s)
	bench_rand_scalar(rng, d, x, n) = (s = 0.0; for _ in 1:n; s += rand(rng, d; control_seq=x); end; s)
	bench_rand_int(rng, d, x, n) = (s = 0; for _ in 1:n; s += rand(rng, d; control_seq=x); end; s)
	bench_rand!_v(rng, d, out, x, n) = (s = 0.0; for _ in 1:n; rand!(rng, d, out; control_seq=x); s += out[1]; end; s)
	bench_rand!_i(rng, d, out, x, n) = (s = 0; for _ in 1:n; rand!(rng, d, out; control_seq=x); s += out[1]; end; s)
	bench_rand_unctrl_scalar(rng, d, n) = (s = 0.0; for _ in 1:n; s += rand(rng, d); end; s)
	bench_rand_unctrl_vec(rng, d, n) = (s = 0.0; for _ in 1:n; s += rand(rng, d)[1]; end; s)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Lines 41 to 43 in 2ae43af

	g = GaussianGLM([0.5, -1.0], 1.0)
	b = BernoulliGLM([0.5, -1.0])
	p = PoissonGLM([0.5, -1.0])

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 47 in 2ae43af

yv = [0.1, 0.2]; yi = [0, 1]

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Lines 50 to 51 in 2ae43af

	bench_logd(g, 0.5, x, 1); bench_logd(b, 1, x, 1); bench_logd(p, 2, x, 1)
	bench_logd(mg, yv, x, 1); bench_logd(mb, yi, x, 1); bench_logd(mp, yi, x, 1)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Lines 54 to 58 in 2ae43af

	@test (@allocated bench_logd(g, 0.5, x, REPS)) == 0
	@test (@allocated bench_logd(b, 1, x, REPS)) == 0
	@test (@allocated bench_logd(p, 2, x, REPS)) == 0
	@test (@allocated bench_logd(mb, yi, x, REPS)) == 0
	@test (@allocated bench_logd(mp, yi, x, REPS)) == 0

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Lines 68 to 70 in 2ae43af

	g = GaussianGLM([0.5, -1.0], 1.0)
	b = BernoulliGLM([0.5, -1.0])
	p = PoissonGLM([0.5, -1.0])

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 101 in 2ae43af

gg = GaussianGLM([0.0, 0.0], 1.0); fit!(gg, yg, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 116 in 2ae43af

gb = BernoulliGLM(zeros(2)); fit!(gb, yb, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 121 in 2ae43af

gp = PoissonGLM(zeros(2)); fit!(gp, yp, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 128 in 2ae43af

gmb = MvBernoulliGLM(zeros(2, 2)); fit!(gmb, ymb, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 133 in 2ae43af

gmp = MvPoissonGLM(zeros(2, 2)); fit!(gmp, ymp, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 141 in 2ae43af

bench_logd_unctrl(zip, 0, 1); bench_logd_unctrl(zip, 5, 1)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/allocations.jl

Line 151 in 2ae43af

zip2 = PoissonZeroInflated(1.0, 0.1); fit!(zip2, y, w)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 12 in 2ae43af

function _synthetic_gaussian_glm(rng, n::Int, p::Int; β::Vector{Float64}, σ2::Float64, weights=:uniform)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 14 in 2ae43af

X = hcat(ones(n), randn(rng, n, p-1)) # intercept + random features

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 17 to 18 in 2ae43af

	w = weights === :uniform ? ones(n) :
	weights === :random ? (rand(rng, n) .+ 0.5) :

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 20 in 2ae43af

return X, y, w

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 26 to 34 in 2ae43af

	@testset "Constructor" begin
	glm = GaussianGLM([0.5, -1.0], 1.0)
	@test glm.β == [0.5, -1.0]
	@test glm.σ2 == 1.0
	glm2 = GaussianGLM([1, 2], 3)
	@test glm2.β == [1, 2]
	@test glm2.σ2 == 3
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 36 to 37 in 2ae43af

	@testset "DensityInterface" begin
	glm = GaussianGLM([0.5, -1.0], 1.0)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 39 to 40 in 2ae43af

	# Trait
	@test DensityKind(glm) == HasDensity()

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 42 to 44 in 2ae43af

	X = [1.0 2.0; 1.0 3.0; 1.0 4.0]
	μ = X * glm.β
	y = rand.(Ref(rng), Normal.(μ, sqrt(glm.σ2)))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 46 to 50 in 2ae43af

	# logdensityof matches closed form
	for i in eachindex(y)
	x_i = vec(X[i, :]) # ensure Vector
	logp = logdensityof(glm, y[i]; control_seq=x_i)
	@test isfinite(logp)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 52 to 54 in 2ae43af

	expected = _expected_logpdf(y[i], dot(glm.β, x_i), glm.σ2)
	@test logp ≈ expected rtol=1e-12 atol=0.0
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 56 to 60 in 2ae43af

	# Higher density near the mean than far away
	x0 = vec(X[1, :])
	y_at_mean = dot(glm.β, x0)
	y_far = y_at_mean + 10.0
	@test logdensityof(glm, y_at_mean; control_seq=x0) > logdensityof(glm, y_far; control_seq=x0)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 63 to 67 in 2ae43af

	@testset "Random sampling" begin
	glm = GaussianGLM([0.5, -1.0], 2.0) # variance 2.0
	x = [1.0, 3.0]
	n = 10_000
	samples = [rand(rng, glm; control_seq=x) for _ in 1:n]

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 69 in 2ae43af

@test all(s -> s isa Real, samples)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 71 to 74 in 2ae43af

	# Empirical mean should be close to dot(β,x)
	μ = dot(glm.β, x)
	m = mean(samples)
	@test m ≈ μ atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 76 to 79 in 2ae43af

	# Empirical variance should be close to σ2
	v = var(samples)
	@test v ≈ glm.σ2 atol=0.25
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 81 to 84 in 2ae43af

	@testset "fit! with uniform weights" begin
	β_true = [0.3, -1.2] # p=2
	σ2_true = 0.7
	X, y, w = _synthetic_gaussian_glm(rng, 2000, 2; β=β_true, σ2=σ2_true, weights=:uniform)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 86 to 87 in 2ae43af

	glm = GaussianGLM([0.0, 0.0], 1.0)
	fit!(glm, y, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 89 to 91 in 2ae43af

	# Coefficients should be close with enough data
	@test glm.β ≈ β_true atol=0.08
	@test glm.σ2 ≈ σ2_true atol=0.10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 93 to 96 in 2ae43af

	@test all(isfinite, glm.β)
	@test isfinite(glm.σ2)
	@test glm.σ2 > 0
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 98 to 101 in 2ae43af

	@testset "fit! with weighted observations" begin
	β_true = [1.0, 0.6]
	σ2_true = 1.5
	X, y, w = _synthetic_gaussian_glm(rng, 2500, 2; β=β_true, σ2=σ2_true, weights=:random)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 103 to 104 in 2ae43af

	glm = GaussianGLM([0.0, 0.0], 1.0)
	fit!(glm, y, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 106 to 108 in 2ae43af

	@test glm.β ≈ β_true atol=0.10
	@test glm.σ2 ≈ σ2_true atol=0.15
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 110 to 112 in 2ae43af

	@testset "Constructor with prior" begin
	glm = GaussianGLM([0.0, 0.0], 1.0)
	@test glm.prior isa NoPrior

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 114 to 117 in 2ae43af

	glm2 = GaussianGLM([0.0, 0.0], 1.0, RidgePrior(2.0))
	@test glm2.prior isa RidgePrior
	@test glm2.prior.λ == 2.0
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 119 to 122 in 2ae43af

	@testset "fit! with RidgePrior shrinks toward zero" begin
	β_true = [3.0, -3.0]
	σ2_true = 1.0
	X, y, w = _synthetic_gaussian_glm(rng, 300, 2; β=β_true, σ2=σ2_true)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 124 to 125 in 2ae43af

	glm_noprior = GaussianGLM([0.0, 0.0], 1.0)
	fit!(glm_noprior, y, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 127 to 128 in 2ae43af

	glm_ridge = GaussianGLM([0.0, 0.0], 1.0, RidgePrior(10.0))
	fit!(glm_ridge, y, w; control_seq=X)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 130 to 132 in 2ae43af

	@test norm(glm_ridge.β) < norm(glm_noprior.β)
	@test all(isfinite, glm_ridge.β)
	end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 134 in 2ae43af

end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 136 to 138 in 2ae43af

	function _synthetic_mvgaussian_glm(rng, n::Int, p::Int, k::Int;
	B::Matrix{Float64}, Σ::Matrix{Float64},
	weights=:uniform)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 141 in 2ae43af

X = hcat(ones(n), randn(rng, n, p-1))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 148 to 149 in 2ae43af

	w = weights === :uniform ? ones(n) :
	weights === :random ? (rand(rng, n) .+ 0.5) :

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 151 in 2ae43af

return X, obs_seq, w

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 174 in 2ae43af

@test_throws DimensionMismatch MvGaussianGLM(B, [1.0 0.0 0.0; 0.0 1.0 0.0; 0.0 0.0 1.0])

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 192 in 2ae43af

@test logdensityof(glm, y; control_seq=x) ≈ expected rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 196 in 2ae43af

logdensityof(glm, μ .+ 10.0; control_seq=x)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 214 in 2ae43af

@test m ≈ μ_true atol=0.05

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 217 in 2ae43af

@test S ≈ Σ atol=0.1

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 223 to 224 in 2ae43af

	X, obs_seq, w = _synthetic_mvgaussian_glm(rng, 4000, 3, 2;
	B=B_true, Σ=Σ_true)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 229 to 230 in 2ae43af

	@test glm.B ≈ B_true atol=0.08
	@test glm.Σ ≈ Σ_true atol=0.10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 239 to 241 in 2ae43af

	X, obs_seq, w = _synthetic_mvgaussian_glm(rng, 3000, 2, 2;
	B=B_true, Σ=Σ_true,
	weights=:random)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 246 to 247 in 2ae43af

	@test glm.B ≈ B_true atol=0.10
	@test glm.Σ ≈ Σ_true atol=0.12

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 253 to 254 in 2ae43af

	X, obs_seq, w = _synthetic_mvgaussian_glm(rng, 200, 2, 2;
	B=B_true, Σ=Σ_true)

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 278 in 2ae43af

end

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 9 in 2ae43af

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 95 to 96 in 2ae43af

	@test logdensityof(glm, 1; control_seq=x) ≈ log(μ) rtol=1e-10
	@test logdensityof(glm, 0; control_seq=x) ≈ log(1 - μ) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 119 in 2ae43af

@test mean(samples) ≈ _sigmoid(dot(glm.β, x)) atol=0.05

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 129 in 2ae43af

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 140 in 2ae43af

@test glm.β ≈ β_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 163 in 2ae43af

@test_throws DimensionMismatch fit!(glm, ones(Int, 5), ones(5); control_seq=ones(5, 3))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 180 in 2ae43af

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 202 in 2ae43af

@test logdensityof(glm, k; control_seq=x) ≈ logpdf(Poisson(μ), k) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 215 in 2ae43af

@test mean(samples) ≈ exp(glm.β[1]) atol=0.2

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 225 in 2ae43af

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 236 in 2ae43af

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 258 in 2ae43af

@test_throws DimensionMismatch fit!(glm, ones(Int, 5), ones(5); control_seq=ones(5, 3))

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 280 to 281 in 2ae43af

	obs_seq[i] = Int[rand(rng) < _sigmoid(dot(B_true[:, j], X[i, :])) ? 1 : 0
	for j in 1:k]

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 326 to 329 in 2ae43af

	@test logdensityof(glm, [1, 1]; control_seq=x) ≈ log(μ1) + log(μ2) rtol=1e-10
	@test logdensityof(glm, [1, 0]; control_seq=x) ≈ log(μ1) + log(1 - μ2) rtol=1e-10
	@test logdensityof(glm, [0, 1]; control_seq=x) ≈ log(1 - μ1) + log(μ2) rtol=1e-10
	@test logdensityof(glm, [0, 0]; control_seq=x) ≈ log(1 - μ1) + log(1 - μ2) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 350 to 351 in 2ae43af

	@test mean(Y[:, 1]) ≈ _sigmoid(dot(B[:, 1], x)) atol=0.05
	@test mean(Y[:, 2]) ≈ _sigmoid(dot(B[:, 2], x)) atol=0.05

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 361 in 2ae43af

@test glm.B ≈ B_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 372 in 2ae43af

@test glm.B ≈ B_true atol=0.25

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 400 in 2ae43af

@test glm.B[:, j] ≈ glm_j.β rtol=1e-6

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 409 in 2ae43af

glm, [zeros(Int, 2) for _ in 1:4], ones(5); control_seq=X,

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 448 in 2ae43af

@test logdensityof(glm, [k1, k2]; control_seq=x) ≈ expected rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 465 to 466 in 2ae43af

	@test mean(Y[:, 1]) ≈ exp(B[1, 1]) atol=0.2
	@test mean(Y[:, 2]) ≈ exp(B[1, 2]) atol=0.2

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 476 in 2ae43af

@test glm.B ≈ B_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 487 in 2ae43af

@test glm.B ≈ B_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 515 in 2ae43af

@test glm.B[:, j] ≈ glm_j.β rtol=1e-6

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 524 in 2ae43af

glm, [zeros(Int, 2) for _ in 1:4], ones(5); control_seq=X,

[codecov]

Codecov Report

❌ Patch coverage is 99.35170% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 99.24%. Comparing base (11a0847) to head (941714f).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
src/glms/glm.jl	99.59%	2 Missing ⚠️
src/multivariate/t.jl	98.42%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #17      +/-   ##
==========================================
+ Coverage   98.83%   99.24%   +0.41%     
==========================================
  Files           2        3       +1     
  Lines         257      796     +539     
==========================================
+ Hits          254      790     +536     
- Misses          3        6       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

JuliaFormatter

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Line 233 in 09348a0

@test S ≈ Σ atol=0.1

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 244 to 245 in 09348a0

	@test glm.B ≈ B_true atol=0.08
	@test glm.Σ ≈ Σ_true atol=0.10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/gaussian.jl

Lines 261 to 262 in 09348a0

	@test glm.B ≈ B_true atol=0.10
	@test glm.Σ ≈ Σ_true atol=0.12

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 94 to 95 in 09348a0

	@test logdensityof(glm, 1; control_seq=x) ≈ log(μ) rtol=1e-10
	@test logdensityof(glm, 0; control_seq=x) ≈ log(1 - μ) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 118 in 09348a0

@test mean(samples) ≈ _sigmoid(dot(glm.β, x)) atol=0.05

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 128 in 09348a0

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 139 in 09348a0

@test glm.β ≈ β_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 202 in 09348a0

@test logdensityof(glm, k; control_seq=x) ≈ logpdf(Poisson(μ), k) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 215 in 09348a0

@test mean(samples) ≈ exp(glm.β[1]) atol=0.2

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 225 in 09348a0

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 236 in 09348a0

@test glm.β ≈ β_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 329 to 332 in 09348a0

	@test logdensityof(glm, [1, 1]; control_seq=x) ≈ log(μ1) + log(μ2) rtol=1e-10
	@test logdensityof(glm, [1, 0]; control_seq=x) ≈ log(μ1) + log(1 - μ2) rtol=1e-10
	@test logdensityof(glm, [0, 1]; control_seq=x) ≈ log(1 - μ1) + log(μ2) rtol=1e-10
	@test logdensityof(glm, [0, 0]; control_seq=x) ≈ log(1 - μ1) + log(1 - μ2) rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 353 to 354 in 09348a0

	@test mean(Y[:, 1]) ≈ _sigmoid(dot(B[:, 1], x)) atol=0.05
	@test mean(Y[:, 2]) ≈ _sigmoid(dot(B[:, 2], x)) atol=0.05

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 364 in 09348a0

@test glm.B ≈ B_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 375 in 09348a0

@test glm.B ≈ B_true atol=0.25

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 403 in 09348a0

@test glm.B[:, j] ≈ glm_j.β rtol=1e-6

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 451 in 09348a0

@test logdensityof(glm, [k1, k2]; control_seq=x) ≈ expected rtol=1e-10

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Lines 468 to 469 in 09348a0

	@test mean(Y[:, 1]) ≈ exp(B[1, 1]) atol=0.2
	@test mean(Y[:, 2]) ≈ exp(B[1, 2]) atol=0.2

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 479 in 09348a0

@test glm.B ≈ B_true atol=0.15

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 490 in 09348a0

@test glm.B ≈ B_true atol=0.20

---

[JuliaFormatter] _{reported by reviewdog 🐶}

EmissionModels.jl/test/glm/test_bernoulli_poisson.jl

Line 518 in 09348a0

@test glm.B[:, j] ≈ glm_j.β rtol=1e-6