Skip to contents

Configure scoring metrics for the models

Source: R/DiseasyScoring.R
DiseasyScoring.Rd

The DiseasyScoring module is responsible for defining the scoring rules (metrics) for comparing observations with model trajectories.

Value

A new instance of the DiseasyScoring R6 class.

See also

DiseasyObservables

Super class

DiseasyBaseModule -> DiseasyScoring

Active bindings

noise_model

(list)
The noise_model currently in the module. Read-only.

observables

(diseasy::DiseasyObservables)
The local copy of an DiseasyObservables module. Read-only.

Methods

Public methods

Inherited methods

DiseasyScoring$new()

Creates a new instance of the DiseasyScoring R6 class.

Usage

DiseasyScoring$new(...)

Arguments

...

parameters sent to DiseasyBaseModule R6 constructor.

DiseasyScoring$set_noise_model()

Sets the DiseasyScoring module to use the specified noise model for the observations.

Usage

DiseasyScoring$set_noise_model(
  observation_noise = c("poisson", "negative binomial"),
  overdispersion = NULL
)

Arguments

observation_noise

(character(1))
The assumed distribution of the observation noise.

overdispersion

(numeric(1))
The assumed overdispersion the observation noise (only for "negative binomial").

Returns

NULL (called for side effects)

DiseasyScoring$compute_likelihood()

Compute the likelihood for a given set of model trajectories

Usage

DiseasyScoring$compute_likelihood(model_trajectories, stratification = NULL)

Arguments

model_trajectories

(data.frame(1))
Model trajectories returned by $get_results() from a model.

stratification

(list(quosures) or NULL)
Use rlang::quos(...) to specify stratification. If given, expressions in stratification evaluated to give the stratification level.

Returns

(data.frame(1))
The likelihood of each realisation in model_trajectories

DiseasyScoring$describe()

Prints a human readable report of the internal state of the module.

Usage

DiseasyScoring$describe()

DiseasyScoring$clone()

The objects of this class are cloneable with this method.

Usage

DiseasyScoring$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

  # Create a scoring module
  scoring <- DiseasyScoring$new()

  # By default, observation noise is assumed to be negative binomial
  scoring
#> # DiseasyScoring #############################################
#> Observation noise model: poisson

  # Generate model trajectory to score
  conn <- DBI::dbConnect(duckdb::duckdb(), ":memory:")
#> duckdb keeps downloaded extensions and secrets in a temporary directory:
#> ℹ /tmp/Rtmp3L2qcp/duckdb
#> This is removed when the R session ends.
#> • Extensions are re-downloaded each session.
#> • Secrets are lost.
#> ℹ Run duckdb(shared_home = TRUE) (or create ~/.duckdb) to keep them (suitable for most users).
#> ℹ Run duckdb(shared_home = FALSE) to accept the temporary directory (and silence this message).
#> ℹ See ?duckdb_storage for details and alternatives.
  observables <- DiseasyObservables$new(
    diseasystore = "Simulist",
    conn = conn
  )
  observables$set_last_queryable_date(as.Date("2020-03-01"))

  model <- DiseasyModelG0$new(
    observables = observables,
    scoring = scoring
  )

  model$plot(observable = "n_admission", prediction_length = 10)


  trajectories <- model$get_results(observable = "n_admission", prediction_length = 10)

  # The scoring of the model trajectories are done with the "$compute_likelihood()" method:
  model$scoring$compute_likelihood(trajectories)
#> # A tibble: 100 × 3
#>    realisation_id weight negative_log_likelihood
#>    <chr>           <dbl>                   <dbl>
#>  1 1                   1                    63.4
#>  2 10                  1                    65.1
#>  3 100                 1                    64.4
#>  4 11                  1                    64.5
#>  5 12                  1                    62.5
#>  6 13                  1                    63.0
#>  7 14                  1                    57.5
#>  8 15                  1                    58.4
#>  9 16                  1                    64.0
#> 10 17                  1                    64.1
#> # ℹ 90 more rows

  DBI::dbDisconnect(conn)
  rm(scoring, conn)