Skip to contents

fittr

fittr provides a framework for fitting many formula-based models in parallel across multiple datasets and dependent variables.

In applied research, a typical project involves dozens of related regressions: multiple outcome variables, several model specifications, subsamples, and weighted/unweighted variants. fittr lets you declare your models, datasets, and definitions once, then generates every combination and runs them — optionally in parallel — returning a cleanly named nested list of fitted model objects.

Any modelling function that accepts formula, data, and weights arguments is supported: lm(), glm(), mixed-effects models, survival models, GAMs, and more.

Installation 

# Install from Codeberg
pak::pak("codeberg::sam_crawley/fittr")

Usage 

library(fittr)

# Define model specifications
base <- model_spec(cyl, disp)

models <- list(
  base     = base,
  extended = base |> add_terms(hp, wt),
  reduced  = base |> drop_terms(disp)
)

# Define dependent variables
defs <- list(
  mpg  = list(dv = "mpg",  fun = "lm"),
  qsec = list(dv = "qsec", fun = "lm")
)

# Preview what will run
preview_fit(models, mtcars, defs)

# Fit all combinations
results <- fit_models(models, mtcars, defs)

# Results are standard model objects
coef(results$mpg$extended)
summary(results$qsec$base)

Key features

  • Multiple model specifications built compositionally from a shared base with model_spec(), add_terms(), drop_terms(), and other helpers.
  • Multiple datasets — pass a named list of data frames; each model runs on every dataset.
  • Weighted regressions — supply weight_var to run weighted and unweighted variants automatically.
  • Per-DV customisation — add extra terms, additional models, or modify specific models for individual dependent variables.
  • Model-specific functions — use different estimators for different models within the same DV.
  • Models that need more than a formula — attach arguments for the fitting function (set_args()) and formula sections after |, such as fixest’s fixed effects (add_part()).
  • Parallel execution — powered by furrr/future, enabled with a single workers argument.
  • Preview before fitting — preview_fit() shows every run as a tibble so you can verify formulas before committing.

Learn more

See vignette("fittr") for a detailed walkthrough.