tabular: Compute complex table

Description Usage Arguments Details Value Formatting tabular methods Author(s) References See Also Examples


Computes a table of summary statistics, cross-classified by various variables.


tabular(table, ...)
## Default S3 method:
tabular(table, ...)
## S3 method for class 'formula'
tabular(table, data = NULL, n, suppressLabels = 0, ...)
## S3 method for class 'tabular'
print(x, justification="n", ...)
## S3 method for class 'tabular'
format(x, digits=4, justification="n", latex=FALSE, html=FALSE, 
                         leftpad = TRUE, rightpad = TRUE, minus = TRUE, 
			 mathmode = TRUE, ...)
## S3 method for class 'tabular'
x[i, j, ..., drop=FALSE]
## S3 method for class 'tabular'
cbind(..., deparse.level = 1)
## S3 method for class 'tabular'
rbind(..., deparse.level = 1)



A table expression. See the Details below.


An optional dataframe, list or environment in which to look for variables in the table.


An optional value giving the length of the data. See the Details below.


How many initial labels to suppress?


The object to print, format, or subset.

digits, ...

In the print and format methods, how many significant digits or other parameters to show by default? See Formatting below.


The default justification to use in the table.


If TRUE, the latexNumeric function will be applied when formatting numeric columns after format, to maintain spacing and handle signs properly.


If TRUE, the htmlNumeric function will be applied when formatting numeric columns after format, to maintain spacing and handle signs properly.

leftpad, rightpad, minus, mathmode

Options to pass to latexNumeric or htmlNumeric to control details of formatting. See those pages for details.

i, j, drop

The usual arguments for matrix indexing, but see the Details below.


Ignored. (Present because the generic requires it.)


For the purposes of this function, a "table" is a rectangular array of values, computed using a formula expression. The left hand side of the formula describes the rows of the table, the right hand side describes the columns.

Within the expression for the rows or columns, the operators +, * and = have special meanings.

The + operator represents concatenation, so that x + y ~ z says to show the rows corresponding to x above the rows corresponding to y.

The * operator represents nesting, so that x*y ~ z says to show the rows of y within each row corresponding to x.

The = operator sets a new name for a term; it is an abbreviation for the Heading() pseudo-function. (“Pseudo-functions” are described in the tables vignette.) Note that = has low operator precedence and may be confused by the parser with setting function arguments, so parentheses are usually needed.

Parentheses may be used to group terms in the usual arithmetic way, so (x + y)*(u + v) is equivalent to x*u + x*v + y*u + y*v.

The names Format, .Format and Heading have special meaning; see the section on Formatting below.

The interpretation of other terms in the formulas depends on how they evaluate.

If the term evaluates to a function, it should be a summary function that produces a scalar value when applied to a vector of values, and that scalar will be displayed in the table. For example, (mean + var) ~ x will display the mean of x above the variance of x. If no function is specified, length is assumed, so the table will display counts. (At most one summary function may be specified in any one term, so mean*var would be an error.)

If the term evaluates to a logical vector, it is assumed to specify a subset. For example, ~ (x > 3) + (x > 5) will tabulate the counts of those two subsets.

If the term evaluates to a factor, it generates multiple rows or columns, corresponding to the different levels of the factor. For example if A has three levels, then A ~ mean*x will calculate the mean of x within each level of A.

If the term evaluates to a language object, it is treated as a macro, and expanded in place in the formula.

Other terms are assumed to be R expressions producing a vector of values to be summarized in the table. Only one vector of values can be specified in any given term, but different terms can summarize different values. is.atomic must evaluate to TRUE for these values for them to be recognized.

All logical, factor or other values in the table should be the same length, as if they were columns in a dataframe (but they can be computed values). If n is missing but data is a dataframe, n is set from that. Otherwise, if terms such as 1 appear in a table, the length is assumed to be the same as for previous terms. As a last resort, set the n argument in the call to tabular() explicitly.

The "[" method extracts a subset of the table. For indexing, consider the table to consist of a matrix containing the values. If drop=TRUE, the labels and attributes are dropped. If drop=FALSE, the default, the i and j indices must select a rectangular block of the table; matrix indexing (using a two column matrix or a full matrix of logical values) is not supported.


An object of S3 class "tabular". This is a matrix of mode list, whose entries are computed summary values, with the following attributes:


A matrix of labels for the rows. This will have the same number of rows as the main matrix, but may have multiple columns for different nested levels of labels. If a label covers multiple rows, it is entered in the first row, and NA is used to fill following rows.


Like rowLabels, but labelling the columns.


The original table expression being displayed. A list of the original format specifications are attached as a "fmtlist" attribute.


A matrix of the same shape as the main result, containing NA for default formatting, or an index into the format list.


The tabular() function does no formatting of computed values, but it records requests for formatting. The format.tabular(), print.tabular() and latex.tabular() functions make use of these requests.

By default, columns are formatted using the format function, with arguments digits and any other arguments passed in .... Each column is formatted separately, similarly to how a matrix is printed by default.

To request special formatting, four pseudo-functions are provided. The first is Format. Normally it includes arguments to pass to the format() function, e.g. Format(digits=2). It may instead include a call to a function, e.g. Format(sprintf("%.2f"). In either case the summary values from cells in the table that are nested below the Format specification will be passed to that function in an argument named x, i.e. in the first example, the values would be formatted using format(digits=2, x=values), and in the second, using sprintf("%.2f", x=values). Users can supply their own function to be used for formatting; it should take values in a named argument x and return a character vector of the same length.

This can be used to control formatting in multiple columns at once. For example, Format(digits=2)*(mean + sd) will print both the mean and standard deviation in a single call to format, guaranteeing that the same number of decimal places is used in both. (The iris example below demonstrates this.)

If the latex argument to latex.tabular is TRUE, then an effort is made to retain spacing, and to convert minus signs to the appropriate type of dash using the latexNumeric function.

The second pseudo-function .Format is mainly intended for internal use. It takes a single integer argument, saying that data governed by this call uses the same formatting as another format specification. In this way entries can be commonly formatted even when they are not contiguous. The integers are assigned sequentially as the format specification is parsed; users will likely need trial and error to find the right value in a complicated table with multiple formats.

A third pseudo-function is Justify. It takes character values or names as arguments; how they are treated depends on the output format. In format.tabular, coarse justification is done to left, right or center, using l, r or c respectively. For LaTeX formatting, any string acceptable as a justification string to LaTeX will be passed on.

A final pseudo-function is Heading. Use this function in the row definitions to set a heading on the following column of row labels. (Internally this is how the headings on factor columns are implemented.) If it is called with no argument, it suppresses the following heading. The suppressLabels=n argument to tabular() is equivalent to repeating Heading() n times at the start of the table formula. The = operator is an abbreviation for Heading(); see above.

tabular methods

The default tabular method just applies as.formula to table, and then calls tabular.formula.

The tabular.formula method is the main workhorse of the package. Other authors who wish to produce tables directly from their own structures should normally create a formula whose environment contains all mentioned variables and call tabular.formula with appropriate arguments.


Duncan Murdoch


This function was inspired by my 20 year old memories of the SAS TABULATE procedure.

See Also

table and ftable are base R functions which produce tables of counts. The tables vignette has many more examples and displays the formatted output.


tabular( (Species + 1) ~ (n=1) + Format(digits=2)*
         (Sepal.Length + Sepal.Width)*(mean + sd), data=iris )

# This example shows some of the less common options         
Sex <- factor(sample(c("Male", "Female"), 100, rep=TRUE))
Status <- factor(sample(c("low", "medium", "high"), 100, rep=TRUE))
z <- rnorm(100)+5
fmt <- function(x) {
  s <- format(x, digits=2)
  even <- ((1:length(s)) %% 2) == 0
  s[even] <- sprintf("(%s)", s[even])
tab <- tabular( Justify(c)*Heading()*z*Sex*Heading(Statistic)*Format(fmt())*(mean+sd) 
                ~ Status )
tab[1:2, c(2,3,1)]

Search within the tables package
Search all R packages, documentation and source code

Questions? Problems? Suggestions? or email at

Please suggest features or report bugs with the GitHub issue tracker.

All documentation is copyright its authors; we didn't write any of that.