Generating Fake Data in R

Author

Arvind V

Introduction

Often we need to generate fake data for teaching and demo purposes. This post uncovers several different packages for this purpose.

Set Up the R Packages

library(tidyverse)
library(mosaic)
library(ggformula)
library(vtable)

# Generate Data
# library(simulate) TO BE FOUND AND INSTALLED!!!!
library(regressinator)
library(holodeck)
library(explore)
library(charlatan)
library(ids) # animals, adjectives, sentences, and proquints
library(rcorpora)
library(simstudy)

Using simulate

sim_bernoulli(prob = 0.2, params = NULL, data = df)
sim_beta(shape1 = 0.2, shape2 = 0.8, params = NULL)

Using regressinator

https://www.refsmmat.com/regressinator/

The regressinator is a pedagogical tool for conducting simulations of regression analyses and diagnostics. It can:

  • Simulate populations with predictor variables from arbitrary distributions
  • Simulate response variables that are functions of the predictor variables plus error, or are drawn from a distribution related to the predictors
  • Given a model, simulate from the population sampling distribution of that model’s estimates
  • Given a model fit to data, generate new simulated data based on the model fit
  • Facilitate lineup plots comparing diagnostics on the fitted model to diagnostics where all model assumptions are met.
library(regressinator)

linear_pop <- population(
  x1 = predictor("rnorm", mean = 4, sd = 10),
  x2 = predictor("runif", min = 0, max = 10),
  y = response(
    0.7 + 2.2 * x1 - 0.2 * x2, # relationship between X and Y
    family = gaussian(),       # link function and response distribution
    error_scale = 1.5          # sd; errors are scaled by this amount
  )
)

In general, population() defines a population according to the following relationship:

\[ Y ∼ Some ~ Distribution\\ \] \[ ~g(E[Y | X = x]) = \mu(x)\\ \] \[ where ~ μ(x)=any~function~of~x\\ \]

If family is not specified the default is Gaussian, and the link function g is identity.

We can create a population with binary outcomes and a logistic link function:

logistic_pop <- population(
  x1 = predictor("rnorm", mean = 0, sd = 10),
  x2 = predictor("runif", min = 0, max = 10),
  y = response(0.7 + 2.2 * x1 - 0.2 * x2,
               family = binomial(link = "logit"))
)
Back to top