Summarizing Results
In logitr: Logit Models w/Preference & WTP Space Utility Parameterizations

knitr::opts_chunk$set(
  collapse = TRUE,
  warning = FALSE,
  message = FALSE,
  fig.retina = 3,
  comment = "#>"
)

Extracting summary tables

The most common function used to view an overall summary of a model is the summary() function:

library(logitr)

model <- logitr(
  data    = yogurt,
  outcome = "choice",
  obsID   = "obsID",
  pars    = c("price", "feat", "brand")
)

summary(model)

The summary prints out a table of the model coefficients as well as other information about the model, such as the log-likelihood, the number of observations, etc.

You can extract the coefficient from the summary table as a data.frame using coef(summary(model)):

coefs <- coef(summary(model))
coefs

The {broom} package

Another approach for extracting the model coefficients as a data frame is to use the tidy() function from the {broom} package:

library(broom)

coefs <- tidy(model)
coefs

The tidy() function returns a tibble and provides a more standardized output and interfaces well with other packages. You can also append a confidence interval to the data frame:

coefs <- tidy(model, conf.int = TRUE, conf.level = 0.95)
coefs

Extracting other values

You can also extract other values of interest at the solution, such as:

The estimated coefficients

coef(model)

The estimated standard errors

se(model)

The log-likelihood

logLik(model)

The variance-covariance matrix

vcov(model)

You can also view a summary of statistics about the model using the glance() function from the {broom} package:

glance(model)

Formatted summary tables

The {gtsummary} package

Often times you will need to create summary tables that are formatted for publication. The {gtsummary} package offers a convenient solution that works well with logitrmodels. For example, a formatted summary table can be obtained using the tbl_regression() function:

library(gtsummary)

model |> 
  tbl_regression()

The tbl_regression() function has many options for customizing the output table. For example, you can change the coefficient names with the label argument:

model |> 
  tbl_regression(
    label = list(
        feat = "Newspaper ad shown?",
        brand = "Yogurt's brand"
    )
  )

The {gtsummary} package supports a wide variety of output types, including support for LaTeX. One you create the table with tbl_regression(), you can print it a variety of ways. For example, once you've created a table x,

x <- model |>
  tbl_regression()

you can print it to LaTeX with any of the following ways:

as_gt(x) |> gt::as_latex()
as_kable_extra(x, format = "latex")
as_hux_table(x) |> to_latex()
as_kable(x, format = "latex")

Multiple models can also be printed in the same table:

model1 <- model

model2 <- logitr(
  data    = yogurt,
  outcome = "choice",
  obsID   = "obsID",
  pars    = c("price*feat", "brand")
)

# Make individual tables
t1 <- tbl_regression(model1)
t2 <- tbl_regression(model2)

# Merge tables
tbl_merge(
  tbls = list(t1, t2),
  tab_spanner = c("**Baseline**", "**Interaction**")
)

The {texreg} package

Another option for obtaining a formatted table is to use the {texreg} package. This is particularly useful for obtaining tables formatted for use in LaTeX.

For example, you can print a summary to the screen using screenreg():

library(texreg)

screenreg(model, stars = c(0.01, 0.05, 0.1))

Likewise, you can print the LaTeX code for a summary table using texreg()

library(texreg)

texreg(model, stars = c(0.01, 0.05, 0.1))

Similar to {gtsummary}, multiple models can be printed using screenreg() or texreg():

model1 <- model

model2 <- logitr(
  data    = yogurt,
  outcome = "choice",
  obsID   = "obsID",
  pars    = c("price*feat", "brand")
)

screenreg(
  list(
    model1,
    model2
  ),
  stars = c(0.01, 0.05, 0.1),
  custom.model.names = c("Baseline", "Interaction")
)