Skip to content

datalorax/equatiomatic

Repository files navigation

equatiomatic

lifecycle R build status Codecov test coverage Covrpage Summary

The goal of equatiomatic is to reduce the pain associated with writing LaTeX code from a fitted model. The package aims to support any model supported by broom. See the introduction to equatiomatic for currently supported models.

Installation

Install from CRAN with

install.packages("equatiomatic")

Or get the development version from GitHub with

remotes::install_github("datalorax/equatiomatic")

Basic usage

The gif above shows the basic functionality.

To convert a model to LaTeX, feed a model object to extract_eq():

library(equatiomatic)

# Fit a simple model
mod1 <- lm(mpg ~ cyl + disp, mtcars)

# Give the results to extract_eq
extract_eq(mod1)
#> $$
#> \operatorname{mpg} = \alpha + \beta_{1}(\operatorname{cyl}) + \beta_{2}(\operatorname{disp}) + \epsilon
#> $$

The model can be built in any standard way. It can handle shortcut syntax:

mod2 <- lm(mpg ~ ., mtcars)
extract_eq(mod2)
#> $$
#> \operatorname{mpg} = \alpha + \beta_{1}(\operatorname{cyl}) + \beta_{2}(\operatorname{disp}) + \beta_{3}(\operatorname{hp}) + \beta_{4}(\operatorname{drat}) + \beta_{5}(\operatorname{wt}) + \beta_{6}(\operatorname{qsec}) + \beta_{7}(\operatorname{vs}) + \beta_{8}(\operatorname{am}) + \beta_{9}(\operatorname{gear}) + \beta_{10}(\operatorname{carb}) + \epsilon
#> $$

When using categorical variables, it will include the levels of the variables as subscripts. Here, we use data from the {palmerpenguins} dataset.

mod3 <- lm(body_mass_g ~ bill_length_mm + species, penguins)
extract_eq(mod3)
#> $$
#> \operatorname{body\_mass\_g} = \alpha + \beta_{1}(\operatorname{bill\_length\_mm}) + \beta_{2}(\operatorname{species}_{\operatorname{Chinstrap}}) + \beta_{3}(\operatorname{species}_{\operatorname{Gentoo}}) + \epsilon
#> $$

It helpfully preserves the order the variables are supplied in the formula:

set.seed(8675309)
d <- data.frame(cat1  = rep(letters[1:3], 100),
                cat2  = rep(LETTERS[1:3], each = 100),
                cont1 = rnorm(300, 100, 1),
                cont2 = rnorm(300, 50, 5),
                out   = rnorm(300, 10, 0.5))
mod4 <- lm(out ~ cont1 + cat2 + cont2 + cat1, d)
extract_eq(mod4)
#> $$
#> \operatorname{out} = \alpha + \beta_{1}(\operatorname{cont1}) + \beta_{2}(\operatorname{cat2}_{\operatorname{B}}) + \beta_{3}(\operatorname{cat2}_{\operatorname{C}}) + \beta_{4}(\operatorname{cont2}) + \beta_{5}(\operatorname{cat1}_{\operatorname{b}}) + \beta_{6}(\operatorname{cat1}_{\operatorname{c}}) + \epsilon
#> $$

Appearance

You can wrap the equations so that a specified number of terms appear on the right-hand side of the equation using terms_per_line (defaults to 4):

extract_eq(mod2, wrap = TRUE)
#> $$
#> \begin{aligned}
#> \operatorname{mpg} &= \alpha + \beta_{1}(\operatorname{cyl}) + \beta_{2}(\operatorname{disp}) + \beta_{3}(\operatorname{hp})\ + \\
#> &\quad \beta_{4}(\operatorname{drat}) + \beta_{5}(\operatorname{wt}) + \beta_{6}(\operatorname{qsec}) + \beta_{7}(\operatorname{vs})\ + \\
#> &\quad \beta_{8}(\operatorname{am}) + \beta_{9}(\operatorname{gear}) + \beta_{10}(\operatorname{carb}) + \epsilon
#> \end{aligned}
#> $$

extract_eq(mod2, wrap = TRUE, terms_per_line = 6)
#> $$
#> \begin{aligned}
#> \operatorname{mpg} &= \alpha + \beta_{1}(\operatorname{cyl}) + \beta_{2}(\operatorname{disp}) + \beta_{3}(\operatorname{hp}) + \beta_{4}(\operatorname{drat}) + \beta_{5}(\operatorname{wt})\ + \\
#> &\quad \beta_{6}(\operatorname{qsec}) + \beta_{7}(\operatorname{vs}) + \beta_{8}(\operatorname{am}) + \beta_{9}(\operatorname{gear}) + \beta_{10}(\operatorname{carb}) + \epsilon
#> \end{aligned}
#> $$

When wrapping, you can change whether the lines end with trailing math operators like + (the default), or if they should begin with them using operator_location = "end" or operator_location = "start":

extract_eq(mod2, wrap = TRUE, terms_per_line = 4, operator_location = "start")
#> $$
#> \begin{aligned}
#> \operatorname{mpg} &= \alpha + \beta_{1}(\operatorname{cyl}) + \beta_{2}(\operatorname{disp}) + \beta_{3}(\operatorname{hp})\\
#> &\quad + \beta_{4}(\operatorname{drat}) + \beta_{5}(\operatorname{wt}) + \beta_{6}(\operatorname{qsec}) + \beta_{7}(\operatorname{vs})\\
#> &\quad + \beta_{8}(\operatorname{am}) + \beta_{9}(\operatorname{gear}) + \beta_{10}(\operatorname{carb}) + \epsilon
#> \end{aligned}
#> $$

By default, all text in the equation is wrapped in \operatorname{}. You can optionally have the variables themselves be italicized (i.e. not be wrapped in \operatorname{}) with ital_vars = TRUE:

extract_eq(mod2, wrap = TRUE, ital_vars = TRUE)
#> $$
#> \begin{aligned}
#> mpg &= \alpha + \beta_{1}(cyl) + \beta_{2}(disp) + \beta_{3}(hp)\ + \\
#> &\quad \beta_{4}(drat) + \beta_{5}(wt) + \beta_{6}(qsec) + \beta_{7}(vs)\ + \\
#> &\quad \beta_{8}(am) + \beta_{9}(gear) + \beta_{10}(carb) + \epsilon
#> \end{aligned}
#> $$

R Markdown and previewing

If you include extract_eq() in an R Markdown chunk, knitr will render the equation. If you’d like to see the LaTeX code wrap the call in print().

You can also use the tex_preview() function from the texPreview package to preview the equation in RStudio:

tex_preview(extract_eq(mod1))

Both extract_eq() and tex_preview() work with magrittr pipes, so you can do something like this:

library(magrittr)  # or library(tidyverse) or any other package that exports %>%

extract_eq(mod1) %>%
  tex_preview()

Extra options

There are several extra options you can enable with additional arguments to extract_eq().

Actual coefficients

You can return actual numeric coefficients instead of Greek letters with use_coefs = TRUE:

extract_eq(mod1, use_coefs = TRUE)
#> $$
#> \operatorname{\widehat{mpg}} = 34.66 - 1.59(\operatorname{cyl}) - 0.02(\operatorname{disp})
#> $$

By default, it will remove doubled operators like “+ -”, but you can keep those in (which is often useful for teaching) with fix_signs = FALSE:

extract_eq(mod1, use_coefs = TRUE, fix_signs = FALSE)
#> $$
#> \operatorname{\widehat{mpg}} = 34.66 + -1.59(\operatorname{cyl}) + -0.02(\operatorname{disp})
#> $$

This works in longer wrapped equations:

extract_eq(mod2, wrap = TRUE, terms_per_line = 3,
           use_coefs = TRUE, fix_signs = FALSE)
#> $$
#> \begin{aligned}
#> \operatorname{\widehat{mpg}} &= 12.3 + -0.11(\operatorname{cyl}) + 0.01(\operatorname{disp})\ + \\
#> &\quad -0.02(\operatorname{hp}) + 0.79(\operatorname{drat}) + -3.72(\operatorname{wt})\ + \\
#> &\quad 0.82(\operatorname{qsec}) + 0.32(\operatorname{vs}) + 2.52(\operatorname{am})\ + \\
#> &\quad 0.66(\operatorname{gear}) + -0.2(\operatorname{carb})
#> \end{aligned}
#> $$

Beyond lm()

You’re not limited to just lm models! equatiomatic supports many other models, including logistic regression, probit regression, and ordered logistic regression (with MASS::polr()).

Logistic regression with glm()

model_logit <- glm(sex ~ bill_length_mm + species,
                   data = penguins, family = binomial(link = "logit"))
extract_eq(model_logit, wrap = TRUE, terms_per_line = 3)
#> $$
#> \begin{aligned}
#> \log\left[ \frac { P( \operatorname{sex} = \operatorname{male} ) }{ 1 - P( \operatorname{sex} = \operatorname{male} ) } \right] &= \alpha + \beta_{1}(\operatorname{bill\_length\_mm}) + \beta_{2}(\operatorname{species}_{\operatorname{Chinstrap}})\ + \\
#> &\quad \beta_{3}(\operatorname{species}_{\operatorname{Gentoo}})
#> \end{aligned}
#> $$

Probit regression with glm()

model_probit <- glm(sex ~ bill_length_mm + species,
                    data = penguins, family = binomial(link = "probit"))
extract_eq(model_probit, wrap = TRUE, terms_per_line = 3)
#> $$
#> \begin{aligned}
#> P( \operatorname{sex} = \operatorname{male} ) &= \Phi[\alpha + \beta_{1}(\operatorname{bill\_length\_mm}) + \beta_{2}(\operatorname{species}_{\operatorname{Chinstrap}})\ + \\
#> &\qquad\ \beta_{3}(\operatorname{species}_{\operatorname{Gentoo}})]
#> \end{aligned}
#> $$

Ordered logistic regression with MASS::polr()

set.seed(1234)
df <- data.frame(outcome = factor(rep(LETTERS[1:3], 100),
                                  levels = LETTERS[1:3],
                                  ordered = TRUE),
                 continuous_1 = rnorm(300, 100, 1),
                 continuous_2 = rnorm(300, 50, 5))

model_ologit <- MASS::polr(outcome ~ continuous_1 + continuous_2,
                           data = df, Hess = TRUE, method = "logistic")
model_oprobit <- MASS::polr(outcome ~ continuous_1 + continuous_2,
                            data = df, Hess = TRUE, method = "probit")

extract_eq(model_ologit, wrap = TRUE)
#> $$
#> \begin{aligned}
#> \log\left[ \frac { P( \operatorname{outcome}  \leq  \operatorname{A} ) }{ 1 - P( \operatorname{outcome}  \leq  \operatorname{A} ) } \right] &= \alpha_{1} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2}) \\
#> \log\left[ \frac { P( \operatorname{outcome}  \leq  \operatorname{B} ) }{ 1 - P( \operatorname{outcome}  \leq  \operatorname{B} ) } \right] &= \alpha_{2} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})
#> \end{aligned}
#> $$

extract_eq(model_oprobit, wrap = TRUE)
#> $$
#> \begin{aligned}
#> P( \operatorname{outcome}  \leq  \operatorname{A} ) &= \Phi[\alpha_{1} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})] \\
#> P( \operatorname{outcome}  \leq  \operatorname{B} ) &= \Phi[\alpha_{2} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})]
#> \end{aligned}
#> $$

Ordered regression (logit and probit) with ordinal::clm()

set.seed(1234)
df <- data.frame(outcome = factor(rep(LETTERS[1:3], 100),
                                  levels = LETTERS[1:3],
                                  ordered = TRUE),
                 continuous_1 = rnorm(300, 1, 1),
                 continuous_2 = rnorm(300, 5, 5))

model_ologit <- ordinal::clm(outcome ~ continuous_1 + continuous_2,
                             data = df, link = "logit")
model_oprobit <- ordinal::clm(outcome ~ continuous_1 + continuous_2,
                              data = df, link = "probit")

extract_eq(model_ologit, wrap = TRUE)
#> $$
#> \begin{aligned}
#> \log\left[ \frac { P( \operatorname{outcome}  \leq  \operatorname{A} ) }{ 1 - P( \operatorname{outcome}  \leq  \operatorname{A} ) } \right] &= \alpha_{1} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2}) \\
#> \log\left[ \frac { P( \operatorname{outcome}  \leq  \operatorname{B} ) }{ 1 - P( \operatorname{outcome}  \leq  \operatorname{B} ) } \right] &= \alpha_{2} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})
#> \end{aligned}
#> $$

$$ \begin{aligned} \log\left[ \frac { P( \operatorname{outcome} \leq \operatorname{A} ) }{ 1 - P( \operatorname{outcome} \leq \operatorname{A} ) } \right] &= \alpha_{1} + \beta_{1}(\operatorname{continuous_1}) + \beta_{2}(\operatorname{continuous_2}) \\ \log\left[ \frac { P( \operatorname{outcome} \leq \operatorname{B} ) }{ 1 - P( \operatorname{outcome} \leq \operatorname{B} ) } \right] &= \alpha_{2} + \beta_{1}(\operatorname{continuous_1}) + \beta_{2}(\operatorname{continuous_2}) \end{aligned} $$

extract_eq(model_oprobit, wrap = TRUE)
#> $$
#> \begin{aligned}
#> P( \operatorname{outcome}  \leq  \operatorname{A} ) &= \Phi[\alpha_{1} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})] \\
#> P( \operatorname{outcome}  \leq  \operatorname{B} ) &= \Phi[\alpha_{2} + \beta_{1}(\operatorname{continuous\_1}) + \beta_{2}(\operatorname{continuous\_2})]
#> \end{aligned}
#> $$

$$ \begin{aligned} P( \operatorname{outcome} \leq \operatorname{A} ) &= \Phi[\alpha_{1} + \beta_{1}(\operatorname{continuous_1}) + \beta_{2}(\operatorname{continuous_2})] \\ P( \operatorname{outcome} \leq \operatorname{B} ) &= \Phi[\alpha_{2} + \beta_{1}(\operatorname{continuous_1}) + \beta_{2}(\operatorname{continuous_2})] \end{aligned} $$

Extension

If you would like to contribute to this package, we’d love your help! We are particularly interested in extending to more models. We hope to support any model supported by broom in the future.

Code of Conduct

Please note that the ‘equatiomatic’ project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

A note of appreciation

We’d like to thank the authors of the {palmerpenguins} dataset for generously allowing us to incorporate the penguins dataset in our package for example usage.

Horst AM, Hill AP, Gorman KB (2020). palmerpenguins: Palmer Archipelago (Antarctica) penguin data. R package version 0.1.0. https://allisonhorst.github.io/palmerpenguins/