changed some EstimationResult to FittedSDM (#51)

Author: guilhermebodin

bench/benchmark_estimate.jl

@@ -14,7 +14,7 @@ num_seeds = 3
 @benchmark begin
     gas = GAS($p, $q, $Beta, $scaling)
     opt_method = ScoreDrivenModels.LBFGS(gas, $num_seeds)
-    estimate!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
+    fit!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
 end
 # BenchmarkTools.Trial:
 #   memory estimate:  466.69 MiB
@@ -40,7 +40,7 @@ num_seeds = 3
 @benchmark begin
     gas = GAS($p, $q, $LogNormal, $scaling)
     opt_method = ScoreDrivenModels.LBFGS(gas, $num_seeds)
-    estimate!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
+    fit!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
 end
 # BenchmarkTools.Trial:
 #   memory estimate:  353.02 MiB
@@ -59,7 +59,7 @@ scaling = 0.5
 @benchmark begin
     gas = GAS($p, $q, $LogNormal, $scaling)
     opt_method = ScoreDrivenModels.LBFGS(gas, $num_seeds)
-    estimate!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
+    fit!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
 end
 # BenchmarkTools.Trial:
 #   memory estimate:  8.26 GiB
@@ -77,7 +77,7 @@ scaling = 1.0
 @benchmark begin
     gas = GAS($p, $q, $LogNormal, $scaling)
     opt_method = ScoreDrivenModels.LBFGS(gas, $num_seeds)
-    estimate!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
+    fit!(gas, $simulation; verbose = $verbose, opt_method = opt_method)
 end
 # BenchmarkTools.Trial:
 #   memory estimate:  666.69 MiB

docs/src/examples.md

@@ -13,7 +13,7 @@ y = Vector{Float64}(vec(inflow'))
 gas = GAS([1, 12], [1, 12], LogNormal, 0.0)
 
 # Estimate the model via MLE
-estimate!(gas, y)
+fit!(gas, y)
 
 # Obtain in-sample estimates for the inflow
 y_gas = fitted_mean(gas, y, dynamic_initial_params(y, gas))

examples/garch.jl

@@ -51,15 +51,10 @@ gas = GAS(1, 1, Normal, 1.0, time_varying_params = [2])
 initial_point = [0.0; 0.5; 0.25; 0.75]
 
 # Estimate the model
-res = estimate!(gas, y; 
-                                  initial_params = initial_params, 
-                                  opt_method = IPNewton(gas, [initial_point]; ub = ub, lb = lb))
-
+f = fit(gas, y; initial_params = initial_params, 
+                  opt_method = IPNewton(gas, [initial_point]; ub = ub, lb = lb))
 
 # Another way to estimate the model is to rely on luck and give many random initial_points
 # for a non constrained optimization 
-
-gas = GAS(1, 1, Normal, 1.0, time_varying_params = [2])
-res = estimate!(gas, y; 
-                                  initial_params = initial_params, 
-                                  opt_method = NelderMead(gas, 100))
+f = fit(gas, y; initial_params = initial_params, 
+                  opt_method = NelderMead(gas, 100))

examples/inflow_lognormal.jl

@@ -51,7 +51,7 @@ y = Vector{Float64}(vec(inflow'))
 gas = GAS([1, 12], [1, 12], LogNormal, 0.0)
 
 # Estimate the model via MLE
-estimate!(gas, y)
+fit!(gas, y)
 
 # Obtain in-sample estimates for the inflow
 y_gas = fitted_mean(gas, y, dynamic_initial_params(y, gas))

src/MLE.jl

@@ -1,9 +1,10 @@
-export estimate!
+export fit, fit!
 
 const DEFAULT_INITIAL_PARAM = NaN.*ones(1, 1)
 const DEFAULT_NUM_SEEDS = 3
+const DEFAULT_VERBOSE = 0
 
-struct EstimationResults{T <: AbstractFloat}
+struct FittedSDM{T <: AbstractFloat}
     aic::T
     bic::T
     llk::T
@@ -45,27 +46,30 @@ function update_aux_estimation!(aux_est::AuxEstimation{T}, func::Optim.TwiceDiff
     return
 end
 
-function estimate!(sdm::SDM{D, T}, y::Vector{T};
-                   initial_params::Matrix{T} = DEFAULT_INITIAL_PARAM,
-                   opt_method::AbstractOptimizationMethod = LBFGS(sdm, DEFAULT_NUM_SEEDS),
-                   verbose::Int = 0) where {D, T}
+function fit(sdm::SDM{D, T}, y::Vector{T};
+             initial_params::Matrix{T} = DEFAULT_INITIAL_PARAM,
+             opt_method::AbstractOptimizationMethod = LBFGS(sdm, DEFAULT_NUM_SEEDS),
+             verbose::Int = DEFAULT_VERBOSE) where {D, T}
 
     # Number of seed and number of params to estimate
     n_seeds = length(opt_method.seeds)
     n = length(y)
-
+    
     unknowns = find_unknowns(sdm)
     n_unknowns = length(unknowns)
-
+    
     # Check if the model has no unknowns
     check_model_estimated(n_unknowns) && return sdm
 
+    # Create a copy of the model to estimate
+    sdm_fit = deepcopy(sdm)
+    
     # optimize for each seed
     aux_est = AuxEstimation{T}()
 
     for i = 1:n_seeds
         try 
-            func = TwiceDifferentiable(psi_tilde -> log_lik(psi_tilde, y, sdm, initial_params, unknowns, n), opt_method.seeds[i])
+            func = TwiceDifferentiable(psi_tilde -> log_lik(psi_tilde, y, sdm_fit, initial_params, unknowns, n), opt_method.seeds[i])
             opt_result = optimize(func, opt_method, verbose, i)
             update_aux_estimation!(aux_est, func, opt_result)
             println("seed $i of $n_seeds - $(-opt_result.minimum)")
@@ -91,9 +95,21 @@ function estimate!(sdm::SDM{D, T}, y::Vector{T};
         println(aux_est.opt_result[best_seed])
     end
 
-    # return the estimated 
-    fill_psitilde!(sdm, coefs, unknowns)
-
     println("Finished!")
-    return EstimationResults{T}(aic, bic, best_llk, coefs, num_hessian)
+    return FittedSDM{T}(aic, bic, best_llk, coefs, num_hessian)
+end
+
+function fit!(sdm::SDM{D, T}, y::Vector{T};
+              initial_params::Matrix{T} = DEFAULT_INITIAL_PARAM,
+              opt_method::AbstractOptimizationMethod = LBFGS(sdm, DEFAULT_NUM_SEEDS),
+              verbose::Int = DEFAULT_VERBOSE) where {D, T}
+
+    unknowns = find_unknowns(sdm)
+    # Check if the model has no unknowns
+    n_unknowns = length(unknowns)
+    check_model_estimated(n_unknowns) && return sdm
+    
+    f = fit(sdm, y; initial_params = initial_params, opt_method = opt_method, verbose = verbose)
+    fill_psitilde!(sdm, f.coefs, unknowns)
+    return f
 end

src/ScoreDrivenModels.jl

@@ -3,7 +3,7 @@ module ScoreDrivenModels
 using Distributions, Optim, SpecialFunctions
 using LinearAlgebra
 
-import Base.length
+import Base.length, Base.deepcopy
 
 const SCALINGS  = [0.0, 1/2, 1.0]
 const BIG_NUM = 1e8

src/gas/gas.jl

@@ -7,6 +7,10 @@ mutable struct GAS{D <: Distribution, T <: AbstractFloat} <: SDM{D, T}
     scaling::Real
 end
 
+function deepcopy(gas::GAS{D, T}) where {D, T}
+    return GAS{D, T}(deepcopy(gas.ω), deepcopy(gas.A), deepcopy(gas.B), deepcopy(gas.scaling))
+end
+
 function create_ω(num_params::Int)
     return fill(NaN, num_params)
 end

test/test_estimate.jl

@@ -7,32 +7,32 @@
         @testset "Estimation by passing number of seeds" begin
             # LBFGS
             gas = GAS(1, 1, Beta, 0.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
             # NelderMead
             gas = GAS(1, 1, Beta, 0.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.NelderMead(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.NelderMead(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
             # IPNewton
             gas = GAS(1, 1, Beta, 0.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.IPNewton(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.IPNewton(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
         end
         @testset "Estimation by passing seeds" begin
             # LBFGS
             gas = GAS(1, 1, Beta, 0.0)
             seeds = [[0.1, 0.1, 0.5, 0.5, 0.5, 0.5]]
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, seeds))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, seeds))
             test_coefficients_GAS_1_1(gas, ω, A, B)
             # NelderMead
             gas = GAS(1, 1, Beta, 0.0)
             seeds = [[0.1, 0.1, 0.5, 0.5, 0.5, 0.5]]
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.NelderMead(gas, seeds))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.NelderMead(gas, seeds))
             test_coefficients_GAS_1_1(gas, ω, A, B)
             # IPNewton
             gas = GAS(1, 1, Beta, 0.0)
             seeds = [[0.1, 0.1, 0.5, 0.5, 0.5, 0.5]]
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.IPNewton(gas, seeds))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.IPNewton(gas, seeds))
             test_coefficients_GAS_1_1(gas, ω, A, B)
         end
     end
@@ -44,7 +44,7 @@
             B = [0.5 0; 0 0.5]
             simulation = simulate_GAS_1_1(LogNormal, 0.0, ω, A, B, 13)
             gas = GAS(1, 1, LogNormal, 0.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
         end
         @testset "Scaling = 0.5" begin
@@ -53,7 +53,7 @@
             B = [0.5 0; 0 0.5]
             simulation = simulate_GAS_1_1(LogNormal, 0.5, ω, A, B, 13)
             gas = GAS(1, 1, LogNormal, 0.5)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
         end
         @testset "Scaling = 1.0" begin
@@ -62,13 +62,13 @@
             B = [0.1 0; 0 0.1]
             simulation = simulate_GAS_1_1(LogNormal, 1.0, ω, A, B, 3)
             gas = GAS(1, 1, LogNormal, 1.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
             test_coefficients_GAS_1_1(gas, ω, A, B)
         end
         @testset "GAS([1, 12], [1, 12])" begin
             simulation = simulate_GAS_1_12(LogNormal, 0.0, 123)
             gas = GAS([1, 12], [1, 12], LogNormal, 0.0)
-            estimate!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
+            fit!(gas, simulation; verbose = 1, opt_method = SDM.LBFGS(gas, 3))
             test_coefficients_GAS_1_12(gas)
         end
     end

test/utils.jl

@@ -125,7 +125,7 @@ function test_GARCH_1_1(y, seed::Int, optimizer; atol = 1e-4, rtol = 1e-4)
     ub = [1.0; 1.0; 1.0; 1.0]
     lb = [-1.0; 0.0; 0.0; 0.0]
     gas = GAS(1, 1, Normal, 1.0, time_varying_params = [2])
-    res = estimate!(gas, y; initial_params = ini, verbose = 1, opt_method = optimizer(gas, 10; ub = ub, lb = lb))
+    res = fit!(gas, y; initial_params = ini, verbose = 1, opt_method = optimizer(gas, 10; ub = ub, lb = lb))
 
     @test gas.ω[1] - -0.00616637237701241 ≈ 0 atol = atol rtol = rtol
     @test gas.ω[2] - 0.010760592759725487 ≈ 0 atol = atol rtol = rtol