This package can be downloaded from CRAN or from Github as below
# CRAN installation
install.packages("rmdl")
# Or remote/developmental version
remotes::install_github("shah-in-boots/rmdl")The package rmdl was intended as a way to handle causal- and epidemiology-based modeling by the following principles:
The package is simple to use. The mtcars dataset will serve as the example, and we will use linear regressions as the primary test. This toy example shows that we will be building six models in parallel, with the key exposure being the wt term, and the two outcomes being mpg and hp.
library(rmdl)
#> Loading required package: vctrs
#> Loading required package: tibble
#>
#> Attaching package: 'tibble'
#> The following object is masked from 'package:vctrs':
#>
#> data_frame
f <- fmls(mpg + hp ~ .x(wt) + disp + cyl + am, pattern = "parallel")
m <- fit(f, .fn = lm, data = mtcars, raw = FALSE)
mt <- model_table(mileage = m)
print(mt)
#> <mdl_tbl>
#> id formula_index data_id name model_call formula_call outcome exposure
#> <chr> <list> <chr> <chr> <chr> <chr> <chr> <chr>
#> 1 b4845f79… <dbl [6]> mtcars mile… lm mpg ~ wt + … mpg wt
#> 2 7830bcbe… <dbl [6]> mtcars mile… lm mpg ~ wt + … mpg wt
#> 3 e3cf8431… <dbl [6]> mtcars mile… lm mpg ~ wt + … mpg wt
#> 4 9a66d0e5… <dbl [6]> mtcars mile… lm hp ~ wt + d… hp wt
#> 5 62f89920… <dbl [6]> mtcars mile… lm hp ~ wt + c… hp wt
#> 6 193d7c3e… <dbl [6]> mtcars mile… lm hp ~ wt + am hp wt
#> # ℹ 7 more variables: mediator <chr>, interaction <chr>, strata <lgl>,
#> # level <lgl>, model_parameters <list>, model_summary <list>,
#> # fit_status <lgl>There are several important extended classes that this package introduces, however they are primarily used for internal validation and for shortcuts to allow more effective communication.
fmls are a version of the base R formula object, but contain additional information and have extra featurestm are atomic elements used to describe individual variables, and departs from how terms are generally treated in the {stats} packagemdl and mdl_tbl exist primarily as tidy versions of class regression modelingThe {rmdl} package is intended to be flexible, extensible, and easy-to-use (albeit opinionated). Please see the vignettes for additional information.