Conversion of R regression output to a LaTeX table.
texreg(
l,
file = NULL,
single.row = FALSE,
stars = c(0.001, 0.01, 0.05),
custom.header = NULL,
custom.model.names = NULL,
custom.coef.names = NULL,
custom.coef.map = NULL,
custom.gof.names = NULL,
custom.gof.rows = NULL,
custom.note = NULL,
digits = 2,
leading.zero = TRUE,
symbol = "\\cdot",
override.coef = 0,
override.se = 0,
override.pvalues = 0,
override.ci.low = 0,
override.ci.up = 0,
omit.coef = NULL,
reorder.coef = NULL,
reorder.gof = NULL,
ci.force = FALSE,
ci.force.level = 0.95,
ci.test = 0,
groups = NULL,
custom.columns = NULL,
custom.col.pos = NULL,
bold = 0,
center = TRUE,
caption = "Statistical models",
caption.above = FALSE,
label = "table:coefficients",
booktabs = FALSE,
dcolumn = FALSE,
siunitx = FALSE,
lyx = FALSE,
sideways = FALSE,
longtable = FALSE,
threeparttable = FALSE,
use.packages = TRUE,
table = TRUE,
tabular = TRUE,
no.margin = FALSE,
fontsize = NULL,
scalebox = NULL,
float.pos = "",
...
)A statistical model or a list of statistical models. Lists of
models can be specified as l = list(model.1, model.2, ...).
Different object types can also be mixed.
Using this argument, the resulting table is written to a file
rather than to the R prompt. The file name can be specified as a character
string. Writing a table to a file can be useful for working with MS Office
or LibreOffice. For example, using the htmlreg function, an
HTML table can be written to a file with the extension .doc and
opened with MS Word. The table can then be simply copied into any Word
document, retaining the formatting of the table. Note that LibreOffice can
import only plain HTML; CSS decorations are not supported; the resulting
tables do not retain the full formatting in LibreOffice.
By default, a model parameter takes up two lines of the
table: the standard error is listed in parentheses under the coefficient.
This saves a lot of horizontal space on the page and is the default table
format in most academic journals. If single.row = TRUE is activated,
however, both coefficient and standard error are placed in a single table
cell in the same line.
The significance levels to be used to draw stars. Between 0 and
4 threshold values can be provided as a numeric vector. For example,
stars = numeric(0) will not print any stars and will not print any
note about significance levels below the table. stars = 0.05 will
attach one single star to all coefficients where the p value is below 0.05.
stars = c(0.001, 0.01, 0.05, 0.1) will print one, two, or three
stars, or a symbol as specified by the symbol argument depending on
the p-values.
An optional named list of multi-column headers that are
placed above the model names. For example,
custom.header = list("abc" = 1:3, "ef" = 4:5) will add the label
"abc" to the first three models and "ef" to the fourth and
fifth model. The column with coefficient names and any custom columns added
by the "custom.columns" argument are not counted towards these
positions. If booktabs = TRUE, \cmidrule rules are added
below the respective labels; otherwise \cline lines are used.
A character vector of labels for the models. By
default, the models are named "Model 1", "Model 2", etc. Specifying
model.names = c("My name 1", "My name 2") etc. overrides the default
behavior.
By default, texreg uses the coefficient names
which are stored in the models. The custom.coef.names argument can
be used to replace them by other character strings in the order of
appearance. For example, if a table shows a total of three different
coefficients (including the intercept), the argument
custom.coef.names = c("Intercept", "variable 1", "variable 2") will
replace their names in this order.
Sometimes it happens that the same variable has a different name in different models. In this case, the user can use this function to assign identical names. If possible, the rows will then be merged into a single row unless both rows contain values in the same column.
Where the argument contains an NA value, the original name of the
coefficient is kept. For example, custom.coef.names = c(NA, "age",
NA) will only replace the second coefficient name and leave the first and
third name as they are in the original model.
See also custom.coef.map for an easier and more comprehensive way to
rename, omit, and reorder coefficients.
The custom.coef.map argument can be used to
select, omit, rename, and reorder coefficients.
Users must supply a named list of this form: list("x" = "First
variable", "y" = NA, "z" = "Third variable"). With that particular example
of custom.coef.map,
coefficients will be presented in order: "x", "y",
"z".
variable "x" will appear as "First variable", variable
"y" will appear as "y", and variable "z" will
appear as "Third variable".
all variables not named "x", "y", or "z" will
be omitted from the table.
A character vector which is used to replace the
names of the goodness-of-fit statistics at the bottom of the table. The
vector must have the same length as the number of GOF statistics in the
final table. The argument works like the custom.coef.names argument,
but for the GOF values. NA values can be included where the original
GOF name should be kept.
A named list of vectors for new lines at the
beginning of the GOF block of the table. For example, list("Random
effects" = c("YES", "YES", "NO"), Observations = c(25, 25, 26)) would
insert two new rows into the table, at the beginning of the GOF block
(i.e., after the coefficients). The rows can contain integer, numeric, or
character objects. Note that this argument is processed after the
custom.gof.names argument (meaning custom.gof.names should
not include any of the new GOF rows) and before the reorder.gof
argument (meaning that the new GOF order specified there should contain
values for the new custom GOF rows). Arguments for custom columns are not
affected because they only insert columns into the coefficient block.
With this argument, a replacement text for the
significance note below the table can be provided. If an empty
character object is provided (custom.note = ""), the note
will be omitted completely. If some character string is provided (e.g.,
custom.note = "My note"), the significance legend is replaced by
My note. The original significance legend can be included by
inserting the %stars wildcard. For example, a custom note can be
added right after the significance legend by providing custom.note =
"%stars. My note.".
If the threeparttable argument is used, any note should be preceded
by "\\item", for example
"\\item %stars. \\item Second note. \\item Third note.", and
it is possible to create line breaks in the formatted table by including
"\\\\" and line breaks in the LaTeX code by including
"\n", for example
"\n\\item %stars.\\\\\n\\item Second line.\n".
Set the number of decimal places for coefficients, standard
errors and goodness-of-fit statistics. Do not use negative values! The
argument works like the digits argument in the
round function of the base package.
Most journals require leading zeros of coefficients and
standard errors (for example, 0.35). This is also the default texreg
behavior. Some journals, however, require omission of leading zeros (for
example, .35). This can be achieved by setting leading.zero =
FALSE.
If four threshold values are handed over to the stars
argument, p-values smaller than the largest threshold value but larger than
the second-largest threshold value are denoted by this symbol. The default
symbol is "\\cdot" for the LaTeX dot, "·" for the
HTML dot, or simply "." for the ASCII dot. If the
texreg function is used, any other mathematical LaTeX symbol
or plain text symbol can be used, for example symbol = "\\circ"
for a small circle (note that backslashes must be escaped). If the
htmlreg function is used, any other HTML character or symbol
can be used. For the screenreg function, only plain text characters
can be used.
Set custom values for the coefficients. New coefficients
are provided as a list of numeric vectors. The list contains vectors of
coefficients for each model. There must be as many vectors of coefficients
as there are models. For example, if there are two models with three model
terms each, the argument could be specified as override.coef =
list(c(0.1, 0.2, 0.3), c(0.05, 0.06, 0.07)). If there is only one model,
custom values can be provided as a plain vector (not embedded in a list).
For example: override.coef = c(0.05, 0.06, 0.07).
Set custom values for the standard errors. New standard
errors are provided as a list of numeric vectors. The list contains vectors
of standard errors for each model. There must be as many vectors of
standard errors as there are models. For example, if there are two models
with three coefficients each, the argument could be specified as
override.se = list(c(0.1, 0.2, 0.3), c(0.05, 0.06, 0.07)). If there
is only one model, custom values can be provided as a plain vector (not
embedded in a list).For example: override.se = c(0.05, 0.06, 0.07).
Overriding standard errors can be useful for the implementation of robust
SEs, for example.
Set custom values for the p-values. New p-values are
provided as a list of numeric vectors. The list contains vectors of
p-values for each model. There must be as many vectors of p-values as there
are models. For example, if there are two models with three coefficients
each, the argument could be specified as override.pvalues =
list(c(0.1, 0.2, 0.3), c(0.05, 0.06, 0.07)). If there is only one model,
custom values can be provided as a plain vector (not embedded in a list).
For example: override.pvalues = c(0.05, 0.06, 0.07). Overriding
p-values can be useful for the implementation of robust SEs and p-values,
for example.
Set custom lower confidence interval bounds. This
works like the other override arguments, with one exception: if confidence
intervals are provided here and in the override.ci.up argument, the
standard errors and p-values as well as the ci.force argument are
ignored.
Set custom upper confidence interval bounds. This
works like the other override arguments, with one exception: if confidence
intervals are provided here and in the override.ci.low argument, the
standard errors and p values as well as the ci.force argument are
ignored.
A character string which is used as a regular expression to
remove coefficient rows from the table. For example, omit.coef =
"group" deletes all coefficient rows from the table where the name of the
coefficient contains the character sequence "group". More complex
regular expressions can be used to filter out several kinds of model terms,
for example omit.coef = "(thresh)|(ranef)" to remove all model terms
matching either "thresh" or "ranef". The omit.coef
argument is processed after the custom.coef.names argument, so the
regular expression should refer to the custom coefficient names. To omit
GOF entries instead of coefficient entries, use the custom arguments of the
extract functions instead (see the help entry of the extract
function.
Reorder the rows of the coefficient block of the
resulting table in a custom way. The argument takes a vector of the same
length as the number of coefficients. For example, if there are three
coefficients, reorder.coef = c(3, 2, 1) will put the third
coefficient in the first row and the first coefficient in the third row.
Reordering can be sensible because interaction effects are often added to
the end of the model output although they were specified earlier in the
model formula. Note: Reordering takes place after processing custom
coefficient names and after omitting coefficients, so the
custom.coef.names and omit.coef arguments should follow the
original order.
Reorder the rows of the goodness-of-fit block of the
resulting table in a custom way. The argument takes a vector of the same
length as the number of GOF statistics. For example, if there are three
goodness-of-fit rows, reorder.gof = c(3, 2, 1) will exchange the
first and the third row. Note: Reordering takes place after processing
custom GOF names and after adding new custom GOF rows, so the
custom.gof.names and custom.gof.rows arguments should follow
the original order, and the reorder.gof argument should contain
values for any rows that are added through the custom.gof.rows
argument.
Should confidence intervals be used instead of the default
standard errors and p-values? Most models implemented in the texreg
package report standard errors and p-values by default while few models
report confidence intervals. However, the functions in the texreg
package can convert standard errors and into confidence intervals using
z-scores if desired. To enforce confidence intervals instead of standard
errors, the ci.force argument accepts either a logical value
indicating whether all models or none of the models should be forced to
report confidence intervals (ci.force = TRUE for all and
ci.force = FALSE for none) or a vector of logical values indicating
for each model separately whether the model should be forced to report
confidence intervals (e.g., ci.force = c(FALSE, TRUE, FALSE)).
Confidence intervals are computed using the standard normal distribution
(z-values based on the qnorm function). The
t-distribution is currently not supported because this would require each
extract method to have an additional argument for the degrees
of freedom.
If the ci.force argument is used to convert
standard errors to confidence intervals, what confidence level should be
used? By default, 0.95 is used (i.e., an alpha value of 0.05).
If confidence intervals are reported, the ci.test
argument specifies the reference value to establish whether a
coefficient/CI is significant. The default value ci.test = 0, for
example, will attach a significance star to coefficients if the confidence
interval does not contain 0. A value of ci.test = 1 could be
useful if coefficients are provided on the odds-ratio scale, for example.
If no star should be printed at all, ci.test = NA can be used. It is
possible to provide a single value for all models or a vector with a
separate value for each model. The ci.test argument works both for
models with native support for confidence intervals and in cases where the
ci.force argument is used.
This argument can be used to group the rows of the table into
blocks. For example, there could be one block for hypotheses and another
block for control variables. Each group has a heading, and the row labels
within a group are indented. The partitions must be handed over as a list
of named numeric vectors, where each number is a row index and each name is
the heading of the group. Example: groups = list("first group" = 1:4,
"second group" = 7:8).
An optional list of additional text columns to be
inserted into the coefficient block of the table, for example coefficient
types. The list should contain one or more character vectors with as many
character or numeric elements as there are coefficients/model terms. If the
vectors in the list are named, the names are used as labels in the table
header. For example,
custom.columns = list(type = c("a", "b", "c"), 1:3) will add two
columns; the first one is labeled while the second one is not. Note that
the numeric elements of the second column will be converted to
character objects in this example. The consequence is that decimal
alignment with the dcolumn package is switched off in these columns.
Note that this argument is processed after any arguments that affect the
number of rows.
An optional integer vector of positions for the columns
given in the custom.columns argument. For example, if there are
three custom columns, custom.col.pos = c(1, 3, 3) will insert the
first custom column before the first column of the original table and the
remaining two custom columns after the second column of the original table.
By default, all custom columns are placed after the first column, which
usually contains the coefficient names.
The p-value threshold below which the coefficient shall be
formatted in a bold font. For example, bold = 0.05 will cause all
coefficients that are significant at the 95% level to be formatted in
bold. Note that this is not compatible with the dcolumn or
siunitx arguments in the texreg function. If both
bold and dcolumn or siunitx are TRUE,
dcolumn and siunitx are switched off, and a warning message
appears. Note also that it is advisable to use stars = FALSE
together with the bold argument because having both bolded
coefficients and significance stars usually does not make any sense.
Should the table be horizontally aligned at the center of the page?
Set the caption of the table.
Should the caption of the table be placed above the table? By default, it is placed below the table.
Set the label of the table environment.
Use the booktabs LaTeX package to get thick horizontal rules in the output table (recommended).
Use the dcolumn LaTeX package to get a nice alignment of
the coefficients at the decimal separator (recommended for use with the
texreg function). Note that only one of the three arguments
bold, dcolumn, and siunitx can be used at a time as
they are mutually incompatible.
Use the siunitx LaTeX package to get a nice alignment of
the coefficients at the decimal separator (recommended for use with the
texreg function). Note that only one of the three arguments
bold, dcolumn, and siunitx can be used at a time as
they are mutually incompatible.
logical; if TRUE, each new line in the output is
doubled, which facilitates transferring the output into the LyX document
processor.
If sideways = TRUE is set, the table floating
environment is replaced by a sidewaystable float, and the
rotating package is loaded in the preamble. The argument only has an
effect if table = TRUE is also set.
If longtable = TRUE is set, the longtable
environment from the longtable LaTeX package is used to set tables
across multiple pages. Note that this argument is not compatible with the
sideways and scalebox arguments. These arguments will be
automatically switched off when longtable = TRUE is set.
If threeparttable = TRUE is set, the
threeparttable environment will be used to enclose the
tabular environment in the LaTeX code, and the significance note
will be enclosed in a tablenotes environment. This permits word
wrapping of long table notes and adequate spacing between multiple notes.
See also the custom.note argument. If longtable is used,
the threeparttablex LaTeX package is used instead of the
threeparttable package.
If this argument is set to TRUE (= the default
behavior), the required LaTeX packages are loaded in the beginning. If set
to FALSE, the use package statements are omitted from the output.
By default, texreg puts the actual tabular object
in a table floating environment. To get only the tabular
object without the whole table header, set table = FALSE.
By default, the table contents are wrapped in a tabular
environment. To get only the contents for each row without the environment,
set tabular = FALSE. Note that if tabular = FALSE, the
table argument must also be FALSE, otherwise a warning is
printed. Switching off the tabular environment may be useful for designing
one's own table more flexibly, for example using tabular* or
tabularx environments in LaTeX.
In order to save space, inner margins of tables can be switched off.
The fontsize argument serves to change the font size
used in the table. Valid values are "tiny", "scriptsize",
"footnotesize", "small", "normalsize", "large",
"Large", "LARGE", "huge", and "Huge". Note that
the scalebox argument often achieves better results when the goal is
to change the size of the table.
The scalebox argument serves to resize the table. For
example, scalebox = 1.0 is equivalent to the normal size,
scalebox = 0.5 decreases the size of the table by one half, and
scalebox = 2.0 doubles the space occupied by the table. Note that
the scalebox argument does not work when the longtable
argument is used.
This argument specifies where the table should be located on
the page or in the document. By default, no floating position is specified,
and LaTeX takes care of the position automatically. Possible values include
"h" (here), "p" (page), "t" (top), "b"
(bottom), any combination thereof, e.g., "tb", or any of these
values followed by an exclamation mark, e.g. "t!", in order to
enforce this position. The square brackets do not have to be specified.
A character object with a regression table and LaTeX markup.
The object has an additional "texregTable" class identifier, which
causes the object to be formatted nicely on screen when printed.
The texreg function creates LaTeX code for inclusion in a LaTeX
document or for usage with Sweave or knitr, based on a list of
statistical models.
Leifeld, Philip (2013). texreg: Conversion of Statistical Model Output in R to LaTeX and HTML Tables. Journal of Statistical Software 55(8): 1-24. http://www.jstatsoft.org/v55/i08/.
Other texreg:
htmlreg(),
huxtablereg(),
knitreg(),
matrixreg(),
plotreg(),
screenreg(),
wordreg()
# NOT RUN {
# Linear mixed-effects models
library("nlme")
model.1 <- lme(distance ~ age, data = Orthodont, random = ~ 1)
model.2 <- lme(distance ~ age + Sex, data = Orthodont, random = ~ 1)
texreg(list(model.1, model.2), booktabs = TRUE, dcolumn = TRUE)
# Ordinary least squares model (example from the 'lm' help file)
ctl <- c(4.17,5.58,5.18,6.11,4.50,4.61,5.17,4.53,5.33,5.14)
trt <- c(4.81,4.17,4.41,3.59,5.87,3.83,6.03,4.89,4.32,4.69)
group <- gl(2,10,20, labels = c("Ctl","Trt"))
weight <- c(ctl, trt)
lm.D9 <- lm(weight ~ group)
table.string <- texreg(lm.D9, return.string = TRUE)
cat(table.string)
# }
Run the code above in your browser using DataLab