Prefix method

pipeline/src/EpiAwarePipeline.jl

@@ -12,11 +12,12 @@ module EpiAwarePipeline
 
 using CSV, Dagger, DataFramesMeta, Dates, Distributions, DocStringExtensions, DrWatson,
       EpiAware, Plots, Statistics, ADTypes, AbstractMCMC, Plots, JLD2, MCMCChains, Turing,
-      DynamicPPL, LogExpFunctions, RCall, LinearAlgebra
+      DynamicPPL, LogExpFunctions, RCall, LinearAlgebra, Random
 
 # Exported pipeline types
-export AbstractEpiAwarePipeline, EpiAwarePipeline, RtwithoutRenewalPipeline,
-       RtwithoutRenewalPriorPipeline, EpiAwareExamplePipeline
+export AbstractEpiAwarePipeline, EpiAwarePipeline, AbstractRtwithoutRenewalPipeline,
+       RtwithoutRenewalPriorPipeline, EpiAwareExamplePipeline, SmoothOutbreakPipeline,
+       MeasuresOutbreakPipeline, SmoothEndemicPipeline, RoughEndemicPipeline
 
 # Exported utility functions
 export calculate_processes

pipeline/src/constructors/make_Rt.jl

@@ -19,3 +19,166 @@ function make_Rt(pipeline::AbstractEpiAwarePipeline; A = 0.3, P = 30.0)
                      [1.0 + A * sin.(2 * π * (t - ϕ) / P) for t in 1:(N - length(Rt))]]
     return true_Rt
 end
+
+"""
+Constructs the time-varying reproduction number (Rt) for a given SmoothOutbreakPipeline.
+
+# Arguments
+- `pipeline::SmoothOutbreakPipeline`: The SmoothOutbreakPipeline object for which to construct Rt.
+- `N::Float64 = 70.0`: The number of time steps to consider.
+- `initial_Rt::Float64 = 1.5`: The initial value of Rt.
+- `reduction::Float64 = 0.5`: The reduction factor applied to Rt over time.
+- `r::Float64 = 1 / 3.5`: The rate of change of Rt over time.
+- `shift::Float64 = 35.0`: The time shift applied to Rt.
+
+# Returns
+- `Rt::Vector{Float64}`: The time-varying reproduction number (Rt) as a vector.
+
+# Example
+
+```julia
+using EpiAwarePipeline, Plots
+pipeline = SmoothOutbreakPipeline()
+Rt = make_Rt(pipeline) |> Rt -> plot(Rt,
+    xlabel = "Time",
+    ylabel = "Rt",
+    lab = "",
+    title = "Smooth outbreak scenario")
+```
+"""
+function make_Rt(pipeline::SmoothOutbreakPipeline; N = 70.0, initial_Rt = 1.5,
+        reduction = 0.5, r = 1 / 3.5, shift = 35.0)
+    Rt = map(1.0:N) do t
+        initial_Rt * (1 - reduction * logistic(r * (t - shift)))
+    end
+    return Rt
+end
+
+"""
+Constructs the time-varying reproduction number (Rt) for a given outbreak with measures scenario.
+
+# Arguments
+- `pipeline::MeasuresOutbreakPipeline`: The outbreak pipeline object.
+- `N::Float64`: The total number of time steps.
+- `initial_Rt::Float64`: The initial reproduction number before any measures are implemented.
+- `measures_Rt::Float64`: The reproduction number during the period when measures are implemented.
+- `post_measures_Rt::Float64`: The reproduction number after the measures are lifted.
+- `reduction::Float64`: The eventual reduction factor applied to the reproduction number after measures are lifted over time.
+- `r::Float64`: The rate of decay towards the reduction factor.
+- `shift::Float64`: The time shift applied to the decay function after end of measures.
+- `t₁::Float64`: The time step at which measures are implemented.
+- `t₂::Float64`: The time step at which measures are lifted.
+
+# Returns
+- `Rt::Vector{Float64}`: The time-varying reproduction number.
+
+# Example
+
+```julia
+using EpiAwarePipeline, Plots
+pipeline = MeasuresOutbreakPipeline()
+Rt = make_Rt(pipeline) |> Rt -> plot(Rt,
+    xlabel = "Time",
+    ylabel = "Rt",
+    lab = "",
+    title = "Outbreak scenario with measures")
+```
+"""
+function make_Rt(
+        pipeline::MeasuresOutbreakPipeline; N = 70.0, initial_Rt = 1.5, measures_Rt = 0.8,
+        post_measures_Rt = 1.2, reduction = 0.6, r = 1 / 3.5, shift = 21.0, t₁ = 21.0, t₂ = 42.0)
+    Rt = vcat(map(t -> initial_Rt, 1.0:t₁),
+        map(t -> measures_Rt, (t₁ + 1.0):t₂),
+        map((t₂ + 1):N) do t
+            post_measures_Rt * (1 - reduction * logistic(r * (t - t₂ - shift)))
+        end)
+
+    return Rt
+end
+
+"""
+Constructs the time-varying reproduction number (Rt) for a smoothly varying endemic scenario.
+
+In this context "endemic" means a scenario where the geometric average of the
+reproduction number is 1.0, that is the following holds:
+
+```math
+(\\prod_{t = 1}^n[R_t])^{1/n} = 1.
+```
+
+This implies that the long-term muliplicative growth of the infected community is 1.0.
+
+# Arguments
+- `pipeline::SmoothEndemicPipeline`: The pipeline for which Rt is constructed.
+- `N::Float64 = 70.0`: The number of time points to generate Rt for.
+- `A::Float64 = 0.1`: The amplitude of the sinusoidal variation in Rt.
+- `P::Float64 = 28.0`: The period of the sinusoidal variation in Rt.
+
+# Returns
+- `Rt::Vector{Float64}`: The time-varying reproduction number.
+
+# Examples
+
+```julia
+using EpiAwarePipeline, Plots
+pipeline = SmoothEndemicPipeline()
+Rt = make_Rt(pipeline) |> Rt -> plot(Rt,
+    xlabel = "Time",
+    ylabel = "Rt",
+    lab = "",
+    title = "Smoothly varying endemic scenario")
+```
+"""
+function make_Rt(
+        pipeline::SmoothEndemicPipeline; N = 70.0, A = 0.1, P = 28.0)
+    log_Rt = map(1.0:N) do t
+        A * sinpi(2 * t / P)
+    end
+    Rt = exp.(log_Rt .- mean(log_Rt)) # Normalize to have geometric mean 1.0
+    return Rt
+end
+
+"""
+Constructs the time-varying reproduction number (Rt) for an randomly varying endemic scenario.
+
+In this context "endemic" means a scenario where the geometric average of the
+reproduction number is 1.0, that is the following holds:
+
+```math
+(\\prod_{t = 1}^n[R_t])^{1/n} = 1.
+```
+
+This implies that the long-term muliplicative growth of the infected community is 1.0.
+
+# Arguments
+- `pipeline::RoughEndemicPipeline`: The `RoughEndemicPipeline` object for which Rt is being constructed.
+- `N_steps::Int = 10`: The number of steps to generate random values for Rt.
+- `rng::AbstractRNG = MarsenneTwister(1234)`: The random number generator to use.
+- `stride::Int = 7`: The stride length for repeating each random value.
+- `σ::Float64 = 0.05`: The standard deviation of the random values.
+
+# Returns
+- `Rt::Vector{Float64}`: The time-varying reproduction number Rt.
+
+# Example
+
+```julia
+using EpiAwarePipeline, Plots
+pipeline = RoughEndemicPipeline()
+Rt = make_Rt(pipeline) |> Rt -> plot(Rt,
+    xlabel = "Time",
+    ylabel = "Rt",
+    lab = "",
+    title = "Rough Rt path endemic scenario")
+```
+"""
+function make_Rt(
+        pipeline::RoughEndemicPipeline; N_steps = 10, rng = MersenneTwister(1234), stride = 7, σ = 0.05)
+    X = σ * randn(rng, N_steps)
+    log_Rt = mapreduce(vcat, X) do x
+        fill(x, stride)
+    end
+    Rt = exp.(log_Rt .- mean(log_Rt)) # Normalize to have geometric mean 1.0
+
+    return Rt
+end

pipeline/src/infer/generate_inference_results.jl

@@ -15,16 +15,12 @@ Generate inference results based on the given configuration of inference model o
 """
 function generate_inference_results(
         truthdata, inference_config, pipeline::AbstractEpiAwarePipeline;
-        tspan, inference_method,
-        prefix_name = "observables", datadir_name = "epiaware_observables")
+        tspan, inference_method, datadir_name = "epiaware_observables")
     config = InferenceConfig(
         inference_config; case_data = truthdata["y_t"], tspan, epimethod = inference_method)
 
     # produce or load inference results
-    prfx = prefix_name * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
-           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
-           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
-           string(inference_config["gi_mean"])
+    prfx = _inference_prefix(truthdata, inference_config, pipeline)
 
     inference_results, inferencefile = produce_or_load(
         infer, config, datadir(datadir_name); prefix = prfx)
@@ -51,15 +47,12 @@ which is deleted after the function call.
 """
 function generate_inference_results(
         truthdata, inference_config, pipeline::EpiAwareExamplePipeline;
-        tspan, inference_method, prefix_name = "testmode_observables")
+        tspan, inference_method)
     config = InferenceConfig(inference_config; case_data = truthdata["y_t"],
         tspan = tspan, epimethod = inference_method)
 
     # produce or load inference results
-    prfx = prefix_name * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
-           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
-           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
-           string(inference_config["gi_mean"])
+    prfx = _inference_prefix(truthdata, inference_config, pipeline)
 
     datadir_name = mktempdir()
 
@@ -73,14 +66,12 @@ Method for prior predictive modelling.
 """
 function generate_inference_results(
         inference_config, pipeline::RtwithoutRenewalPriorPipeline;
-        tspan, prefix_name = "prior_observables")
+        tspan)
     config = InferenceConfig(
         inference_config; case_data = missing, tspan, epimethod = DirectSample())
 
     # produce or load inference results
-    prfx = prefix_name * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
-           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
-           string(inference_config["gi_mean"])
+    prfx = _inference_prefix(truthdata, inference_config, pipeline)
 
     datadir_name = mktempdir()
 

pipeline/src/infer/infer.jl

@@ -1,4 +1,5 @@
 include("InferenceConfig.jl")
+include("inference_prefix.jl")
 include("generate_inference_results.jl")
 include("map_inference_results.jl")
 include("define_epiprob.jl")

pipeline/src/infer/inference_prefix.jl

@@ -0,0 +1,54 @@
+"""
+This is an internal method that generates the prefix for the inference results file name.
+"""
+function _inference_prefix(truthdata, inference_config, pipeline::AbstractEpiAwarePipeline)
+    return "observables" * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(truthdata, inference_config, pipeline::EpiAwareExamplePipeline)
+    return "testmode_observables" * "_igp_" * string(inference_config["igp"]) *
+           "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(
+        truthdata, inference_config, pipeline::RtwithoutRenewalPriorPipeline)
+    return "prior_observables" * "_igp_" * string(inference_config["igp"]) *
+           "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(truthdata, inference_config, pipeline::SmoothOutbreakPipeline)
+    return "smooth_outbreak" * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(truthdata, inference_config, pipeline::MeasuresOutbreakPipeline)
+    return "measures_outbreak" * "_igp_" * string(inference_config["igp"]) *
+           "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(truthdata, inference_config, pipeline::SmoothEndemicPipeline)
+    return "smooth_endemic" * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end
+
+function _inference_prefix(truthdata, inference_config, pipeline::RoughEndemicPipeline)
+    return "rough_endemic" * "_igp_" * string(inference_config["igp"]) * "_latentmodel_" *
+           inference_config["latent_namemodels"].first * "_truth_gi_mean_" *
+           string(truthdata["truth_gi_mean"]) * "_used_gi_mean_" *
+           string(inference_config["gi_mean"])
+end

pipeline/src/pipeline/pipelinetypes.jl

@@ -4,11 +4,9 @@ The abstract root type for all pipeline types using `EpiAware`.
 abstract type AbstractEpiAwarePipeline end
 
 """
-The pipeline type for the Rt pipeline with renewal including specific options
-    for plotting and saving.
+An abstract type for different scenarios in the Rt without Renewal pipeline.
 """
-struct RtwithoutRenewalPipeline <: AbstractEpiAwarePipeline
-end
+abstract type AbstractRtwithoutRenewalPipeline <: AbstractEpiAwarePipeline end
 
 """
 The pipeline type for the Rt pipeline without renewal with only prior predictive
@@ -22,3 +20,43 @@ The pipeline type for the Rt pipeline without renewal in test mode.
 """
 struct EpiAwareExamplePipeline <: AbstractEpiAwarePipeline
 end
+
+"""
+The pipeline type for the Rt pipeline for an outbreak scenario where Rt decreases
+    smoothly over time to Rt < 1.
+
+# Example
+
+```julia
+using EpiAwarePipeline, Plots
+pipeline = SmoothOutbreakPipeline()
+Rt = make_Rt(pipeline) |> Rt -> plot(Rt,
+    xlabel = "Time",
+    ylabel = "Rt",
+    lab = "",
+    title = "Smooth outbreak scenario")
+```
+"""
+struct SmoothOutbreakPipeline <: AbstractRtwithoutRenewalPipeline
+end
+
+"""
+The pipeline type for the Rt pipeline for an outbreak scenario where Rt has
+    discontinuous changes over time due to implementation of measures.
+"""
+struct MeasuresOutbreakPipeline <: AbstractRtwithoutRenewalPipeline
+end
+
+"""
+The pipeline type for the Rt pipeline for an endemic scenario where Rt changes in
+    a smooth sinusoidal manner over time.
+"""
+struct SmoothEndemicPipeline <: AbstractRtwithoutRenewalPipeline
+end
+
+"""
+The pipeline type for the Rt pipeline for an endemic scenario where Rt changes in
+    a weekly-varying discontinuous manner over time.
+"""
+struct RoughEndemicPipeline <: AbstractRtwithoutRenewalPipeline
+end

pipeline/src/simulate/generate_truthdata.jl

@@ -21,13 +21,15 @@ the `simulate` function to generate the truth data. This is the default method.
 """
 function generate_truthdata(
         truth_data_config, pipeline::AbstractEpiAwarePipeline; plot = true,
-        datadir_str = "truth_data", prefix = "truth_data")
+        datadir_str = "truth_data")
     default_params = make_default_params(pipeline)
     config = TruthSimulationConfig(
         truth_process = default_params["Rt"], gi_mean = truth_data_config["gi_mean"],
         gi_std = truth_data_config["gi_std"], logit_daily_ascertainment = default_params["logit_daily_ascertainment"],
         cluster_factor = default_params["cluster_factor"], I0 = default_params["I0"])
 
+    prefix = _simulate_prefix(pipeline)
+
     truthdata, truthfile = produce_or_load(
         simulate, config, datadir(datadir_str); prefix = prefix)
     if plot

pipeline/src/simulate/simulate.jl

@@ -1,2 +1,3 @@
 include("TruthSimulationConfig.jl")
+include("simulate_prefix.jl")
 include("generate_truthdata.jl")

pipeline/src/simulate/simulate_prefix.jl

@@ -0,0 +1,12 @@
+"""
+Internal method for setting the prefix for the truth data file name.
+"""
+_simulate_prefix(pipeline::AbstractEpiAwarePipeline) = "truth_data"
+
+_simulate_prefix(pipeline::SmoothOutbreakPipeline) = "truth_data_smooth_outbreak"
+
+_simulate_prefix(pipeline::MeasuresOutbreakPipeline) = "truth_data_measures_outbreak"
+
+_simulate_prefix(pipeline::SmoothEndemicPipeline) = "truth_data_smooth_endemic"
+
+_simulate_prefix(pipeline::RoughEndemicPipeline) = "truth_data_rough_endemic"