DOPE: Generate a distribution of possible effects.

In christophercschwarz/DOPE: Distributions of Possible Effects

Description

Wrapper for parallel simfun. Takes a linear regression model fit by lm and returns the results information on the distribution of possible effects. Currently implimented in both R and C++. The C++ version is faster while the R version is easier for the expected user base to read and modify as needed.

Usage

  DOPE(mod,nsims=10000,language="cpp",n.cores=1,buff=sqrt(.Machine$double.eps))

Arguments

mod	A linear regression model fit by lm.
nsims	numeric. How many draws to take?
language	character, either "cpp" or "R" determining which implimentation to use.
n.cores	numeric. How many cores should the simulation be run on?
buff	numeric. A buffer to avoid numeric positive non-definiteness.

Value

Returns a data.frame of nsims + 1 rows, with the last row containing the input coefficients. Includes intercept, regressor coefficients, control function coefficient, and R-squared as columns.

Examples

  set.seed(8675309)
  x_vars <- 5
  n_obs <- 1000
  corm <- RandomCormCPP(nvars = x_vars)
  X_mat <- MASS::mvrnorm(n_obs, rep(0,x_vars), Sigma = corm, empirical = TRUE)
  
  betas <- 1:x_vars
  
  y <- X_mat 
  
  dat <- data.frame(y,X_mat)
  
  mod <- lm(y ~ ., data=dat)
  dope <- DOPE(mod, nsims = 3000, n.cores = parallel::detectCores())

christophercschwarz/DOPE documentation built on April 10, 2020, 4:43 a.m.