Calculate predicted levels and marginal effects using the delta method to calculate standard errors. This is an R-based version of Stata's 'margins' command.

Features:

• Calculate predictive levels and margins for glm and ivreg objects (more models to be added - PRs welcome) using closed-form derivatives

• Add custom variance-covariance matrices to all calculations to add, e.g., clustered or robust standard errors (for more information on replicating Stata analyses, see here)

• Frequency weights are incorporated into margins and effects

To install this package from CRAN, please run

install.packages('modmarg')

To install the development version of this package, please run

devtools::install_github('anniejw6/modmarg', build_vignettes = TRUE)

Here is an example of estimating predicted levels and effects using the iris dataset:

data(iris)

mod <- glm(Sepal.Length ~ Sepal.Width + Species, 
           data = iris, family = 'gaussian')
           
# Predicted Levels
modmarg::marg(mod, var_interest = 'Species', type = 'levels')

# Predicted Effects
modmarg::marg(mod, var_interest = 'Species', type = 'effects')

There are two vignettes included:

vignette('usage', package = 'modmarg')
vignette('delta-method', package = 'modmarg')

• Delta Method: This is from the appendix the book guide to the MARK program, developed by Gary White.

• How can I estimate the standard error of transformed regression parameters in R using the delta method? from UCLA.

• The Delta method to estimate standard errors from a non-linear transformation from Econometrics by Simulation.

• A FAQ on Margins from the Stata Blog

• Delta Method on Wikipedia

• What is the intuition behind the sandwich estimator? from StackExchange

• Least Squares Optimization by Harald E. Krogstad

• The robust sandwich variance estimator for linear regression (theory) by Jonathan Bartlett

• Using Stata’s Margins Command to Estimate and Interpret Adjusted Predictions and Marginal Effects by Richard Williams