In NErler/simvalidator: Validation of Statistical Models via Simulation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

simvalidator

Work in progress!!!

The goal of simvalidator is to perform simulation analysis to validate model results, and to present these results in Rmarkdown documents.

Installation

# install.packages("remotes")
remotes::install_github("NErler/simvalidator")

Example Workflow

library(simvalidator)

Simulating data

Specify an expression that generates the covariate part of the data, e.g.:

covar_def <- expression(data.frame(sex = factor(rbinom(N, size = 1, prob = 0.5)),
                                   age = rnorm(N, 63, 9))
)

Provide a list of parameters used in the covariate definition and the function that creates the outcome variable:

data_params = list(
  # standard arguments, always required:
  response_type = "lm",
  formula = resp ~ age + sex,
  # arguments for generating covariates (depend on the user-specified expression)
  N = 100,
  # arguments for creating the outcome (depend on the type of outcome)
  reg_coefs = c('(Intercept)' = -50.8, age = 0.7, sex1 = 1.2),
  resid_sd = 0.5)

This list has to include the model formula. All other arguments are specific to the covariate definition and the data simulation function that is used.

The outcome for a single dataset is simulated using the function sim_data:

data <- sim_data(covar_def, data_params, seed = 123)

The seed value will typically be a seed derived from a global_seed (specified in the sim_params). For a simulation study, sim_data is usually called repeatedly internally to create a large number of datasets.

New functions to simulate different types of outcomes should be called sim_outcome_<response_type>.

Fit models

Specify a list of model functions to be used. This list has to include the arguments fun, the function that runs the model, and result, which is used to identify the function used to extract the desired elements of the results from the model.

models <- list(
  set_model(fun = lm, fun_args = NULL,
            result = "default", res_args = list(type = "lm")),
  set_model(fun = JointAI::lm_imp,
            fun_args = list(n.iter = 300),
            result = "JointAI", res_args = NULL)
)

New result functions must be named get_result_<type> where the type is the character string specified in the argument result. get_result_<type> functions have to use the argument fitted_model which takes the fitted model.

With this information the function fit_models is called:

res <- fit_models(models, formula = data_params$formula, data, seed = 123)

This function fits each of the models and applies the corresponding get_result_<type> to the fitted model.

The simulation

sim_params <- list(
  global_seed = 2020,
  nr_sims = 10
)

library("doFuture")
registerDoFuture()

plan(list(multisession, sequential))
sim_out <- run_sim(sim_params, covar_def, data_params, models)

sim_out

NErler/simvalidator documentation built on May 17, 2022, 7:54 a.m.