Create a huxtable to display model output
huxreg(..., error_format = "({std.error})", error_style = c("stderr",
"ci", "statistic", "pvalue"), error_pos = c("below", "same", "right"),
number_format = "%.3f", align = ".", pad_decimal = ".",
ci_level = NULL, tidy_args = NULL, stars = c(`***` = 0.001, `**` =
0.01, `*` = 0.05), bold_signif = NULL, borders = 0.4,
outer_borders = 0.8, note = if (is.null(stars)) NULL else "{stars}.",
statistics = c(N = "nobs", R2 = "r.squared", "logLik", "AIC"),
coefs = NULL, omit_coefs = NULL)Models, or a single list of models. Names will be used as column headings.
How to display uncertainty in estimates. See below.
Deprecated. One or more of "stderr", "ci" (confidence interval), "statistic" or "pvalue".
Display uncertainty "below", to the "right" of, or in the "same" cell as estimates.
Format for numbering. See number_format() for details.
Alignment for table cells. Set to a single character to align on this character.
Deprecated in favour of align.
Confidence level for intervals. Set to NULL to not calculate confidence intervals.
List of arguments to pass to broom::tidy(). You can also pass a list of lists;
if so, the nth element will be used for the nth column.
Levels for p value stars. Names of stars are symbols to use. Set to NULL to not show stars.
Where p values are below this number, cells will be displayed in bold. Use NULL to turn off
this behaviour.
Thickness of inner horizontal borders. Set to 0 for no borders.
Thickness of outer (top and bottom) horizontal borders. Set to 0 for no borders.
Footnote for bottom cell, which spans all columns. {stars} will be replaced by a note about
significance stars. Set to NULL for no footnote.
A vector of summary statistics to display. Set to NULL to show all available
statistics. To change display names, name the statistics vector:
c("Displayed title" = "statistic_name", ...)
A vector of coefficients to display. Overrules omit_coefs. To change display names,
name the coef vector: c("Displayed title" = "coefficient_name", ...)
Omit these coefficients.
A huxtable object.
If you wish to use e.g. robust standard errors, you can pass results from e.g.
lmtest::coeftest() into huxreg, since these objects have tidy methods.
Alternatively, to manually insert your own statistics, see tidy_override().
Models must have a generics::tidy() method defined, which should return "term", "estimate",
"std.error", "statistic" and "p.value". The "broom" package provides methods for many model
objects. If the tidy method does not have a conf.int option,
huxreg will calculate confidence intervals itself, using a normal approximation.
If ... has names or contains a single named list, the names will be used for column headings.
Otherwise column headings will be automatically created.
If the coef and/or statistics vectors have names, these will be used for row headings. If
different values of coef have the same name, the corresponding rows will be merged in the
output.
statistics should be column names from generics::glance(). You can also use "nobs" for the
number of observations. If statistics is NULL then all columns from glance will be used. To
use no columns, set statistics = character(0).
error_format is a string to be interpreted by glue::glue(). Terms in parentheses will be
replaced by computed values. You can use any columns returned
by tidy: typical columns include statistic, p.value, std.error, as well as conf.low
and conf.high if you have set ci_level. For example, to show confidence intervals, you
could write error_format = "{conf.low} to {conf.high}".
# NOT RUN {
if (! requireNamespace("broom")) {
stop("Please install 'broom' to run this example.")
}
lm1 <- lm(mpg ~ cyl, mtcars)
lm2 <- lm(mpg ~ cyl + hp, mtcars)
glm1 <- glm(I(mpg > 20) ~ cyl, mtcars,
family = binomial)
huxreg(lm1, lm2, glm1)
# }
Run the code above in your browser using DataLab