regsimulate: Simulate from a linear regression model

In StatisticsSU/regkurs: Kurspaket - Regressions- och tidsserieanalys

Simulate from a linear regression model

Description

Simulates a dataset with n observation from the linear regression model

y = β_0 + β_1 * x_1 + ... + β_k * x_k + ε, ε ~ N(0, σ_ ε^2)

with covariates (x) simulated from a normal distribution with the same correlation rho_x
between all pairs of covariates. Covariate x_j has standard deviation sigma_x[j].
Alternatively the covariate can follow a uniform distribution.

Usage

regsimulate(
  n,
  betavect,
  sigma_eps,
  intercept = TRUE,
  covdist = "normal",
  rho_x = 0,
  sigma_x = rep(1, length(betavect) - intercept)
)

Arguments

n	the number of observations in the simulated dataset.
betavect	a vector with regression coefficients c(beta_0,beta_1,...beta_k). First element is intercept if intercept = TRUE
sigma_eps	standard deviation of the error terms, epsilon.
intercept	if TRUE an intercept is added to the model.
covdist	distribution of the covariates. Options: 'normal' or 'uniform'.
rho_x	correlation among the covariates. Same for all covariate pairs.
sigma_x	vector with standard deviation of the covariates.

Value

dataframe with simulated data (y, X1, X2, ..., XK) (no intercept included).

Examples

library(regkurs)
simdata <- regsimulate(n = 500, betavect = c(1, -2, 1, 0), sigma_eps = 2)
lmfit <- lm(y ~ X1 + X2 + X3, data = simdata)
regsummary(lmfit, anova = F)

StatisticsSU/regkurs documentation built on Jan. 29, 2023, 4:54 p.m.