Latex Equations From Model Objects the Equatiomatic Package

Walking Through equatiomatic

First get the package installed and loaded:

# package install
# install.packages("equatiomatic", repos = "http://cran.us.r-project.org")

# load up
library(equatiomatic)

I’m using some data from my most recent publication in Public Administration Review, which tests competing theories of body-worn camera (BWC) activation. We ask: Is variation in BWC activations more explained by officer attitudes towards the cameras, by officer demographics, or by job function. I won’t repeat the whole analysis here, but you can find out by visiting the article!

library(tidyverse)

activations <- read_csv("activations.csv")

head(activations)

## # A tibble: 6 x 26
##   totalactivations Activ_Plus_One_~ years_LEO Female  rank BWC_time forcecount
##              <dbl>            <dbl>     <dbl>  <dbl> <dbl>    <dbl>      <dbl>
## 1               54             4.01         4      0     1        5          0
## 2               79             4.38        10      0     1        3          0
## 3              138             4.93         4      0     1        5          1
## 4               11             2.48        11      0     1        5          0
## 5              148             5.00        20      0     1        5          0
## 6              198             5.29         3      0     1        4          1
## # ... with 19 more variables: totalprimarycalls <dbl>, arrests <dbl>,
## #   Line_Officer <dbl>, BWCapproval_new <dbl>, POS_LATENT <dbl>,
## #   BWC_understand <dbl>, BWC_freedom <dbl>, BWC_decision <dbl>,
## #   BWC_manipulate <dbl>, BWC_modify <dbl>, BWC_lessforce <dbl>,
## #   BWC_assault <dbl>, BWC_complaint <dbl>, BWC_personal <dbl>,
## #   BWC_embarrass <dbl>, BWC_hatred <dbl>, BWC_fair <dbl>, BWC_protect <dbl>,
## #   BWC_wellbeing <dbl>

Let’s build a quick (and misspecified here) version of one of the main models of interest in the paper:

job_function <- lm(totalactivations ~ forcecount + totalprimarycalls + arrests + Line_Officer, activations)

Now, we give the results of that model to equatiomatic and let it extract and build:

equatiomatic::extract_eq(job_function,
                         wrap = TRUE,        # Long equation needs to wrap 
                         terms_per_line = 2) # Max two equation terms per line

## $$
## \begin{aligned}
## \operatorname{totalactivations} &= \alpha + \beta_{1}(\operatorname{forcecount})\ + \\
## &\quad \beta_{2}(\operatorname{totalprimarycalls}) + \beta_{3}(\operatorname{arrests})\ + \\
## &\quad \beta_{4}(\operatorname{Line\_Officer}) + \epsilon
## \end{aligned}
## $$

We can take the output directly to Rmarkdown using the given LaTeX!

\[ \begin{aligned} \operatorname{totalactivations} &= \alpha + \beta_{1}(\operatorname{forcecount})\ + \\ &\quad \beta_{2}(\operatorname{totalprimarycalls}) + \beta_{3}(\operatorname{arrests})\ + \\ &\quad \beta_{4}(\operatorname{Line\_Officer}) + \epsilon \end{aligned} \]

Absolutely gorgeous! But it gets better, we can include the coefficients instead of funny Greek letters!

equatiomatic::extract_eq(job_function,
                         use_coefs = TRUE,   # Use coefficients instead of beta
                         wrap = TRUE,        # Long equation needs to wrap 
                         terms_per_line = 2) # Max two equation terms per line

## $$
## \begin{aligned}
## \operatorname{totalactivations} &= 6.19 + 10.66(\operatorname{forcecount})\ + \\
## &\quad 0.65(\operatorname{totalprimarycalls}) + 3.91(\operatorname{arrests})\ + \\
## &\quad 18.78(\operatorname{Line\_Officer}) + \epsilon
## \end{aligned}
## $$

Again, copy/paste over the LaTeX given by equatiomatic, and:

\[ \begin{aligned} \operatorname{totalactivations} &= 6.19 + 10.66(\operatorname{forcecount})\ + \\ &\quad 0.65(\operatorname{totalprimarycalls}) + 3.91(\operatorname{arrests})\ + \\ &\quad 18.78(\operatorname{Line\_Officer}) + \epsilon \end{aligned} \]

By the way, the package isn’t limited to linear regressions, and already has support for logistic and probit regressions with glm(), and ordered logistic regressions. Hit up the package home to follow development.

I am completely impressed by this young package so far, and can’t wait to see what else is coming!