# Creates a cross tabulation with counts and percentages This function is pronounced "presentable"! The original purpose was to create a particular kind of cross tabulation that I ask for in class: counts with column percentages. Requests from users have caused a bit more generality to be built into the function. Now, optionally, it will provide row percents. This is a generic function. Most users will find the formula method most convenient. Use the colpct and rowpct arguments to indicate if column or row percentages are desired. I suggest most users will use the formula method for this. Running a command like this will, generally, do the right thing: `tab <- pctable(y ~ x, data = dat)` There is also a method that will work with characters representing variable names. `tab <- pctable("y", "x", data = dat)` Running the function should write a table in the output console, but it also creates an object (`tab`). That object can be displayed in a number of ways. A summary method is provided, so one could look at different representations of the same table. `summary(tab, rowpct = TRUE, colpct = FALSE)` or `summary(tab, rowpct = TRUE, colpct = TRUE)` Tables that include only row or column percentages will be compatible with the html and latex exporters in the excellent `tables` package. The formula method is the recommended method for users. Run `pctable(myrow ~ mycol, data = dat)`. In an earlier version, I gave different advice, so please adjust your usage. The character method exists only for variety. It accepts character strings rather than a formula to define the columns that should be plotted. The method used most often for most users should be the formula method. ```r pctable(rv, ...) ## Default S3 method: pctable( rv, cv, rvlab = NULL, cvlab = NULL, colpct = TRUE, rowpct = FALSE, rounded = FALSE, ... ) ## S3 method for class 'formula' pctable( formula, data = NULL, rvlab = NULL, cvlab = NULL, colpct = TRUE, rowpct = FALSE, rounded = FALSE, ... ) ## S3 method for class 'character' pctable( rv, cv, data = NULL, rvlab = NULL, cvlab = NULL, colpct = TRUE, rowpct = FALSE, rounded = FALSE, ... ) ``` ## Arguments - `rv`: A row variable name - `...`: Other arguments. So far, the most likely additional arguments are to be passed along to the table function, such as "exclude", "useNA", or "dnn" (which will override the rvlab and cvlab arguments provided by some methods). Some methods will also pass along these arguments to model.frame, "subset" "xlev", "na.action", "drop.unused.levels". - `cv`: Column variable - `rvlab`: Optional: row variable label - `cvlab`: Optional: col variable label - `colpct`: Default TRUE: are column percentags desired in the presentation of this result? - `rowpct`: Default FALSE: are row percentages desired in the presentation of this result - `rounded`: Default FALSE, rounds to 10's for privacy purposes. - `formula`: A two sided formula. - `data`: A data frame. ## Returns A list with tables (count, column percent, row percent) as well as a copy of the call. ## Details Please bear in mind the following. The output object is a list of tables of partial information, which are then assembled in various ways by the print method (print.pctable). A lovely table will appear on the screen, but the thing itself has more information and a less beautiful structure. A print method is supplied. For any `pctable` object, it is possible to run follow-ups like print(tab, rowpct = TRUE, colpct = FALSE) The method `print.pctable(tab)` assembles the object into (my opinion of) a presentable form. The print method has argumnets `rowpct` and `colpct` that determine which percentages are included in the presentation. When using character arguments, the row variable rv rowvar must be a quoted string if the user intends the method pctable.character to be dispatched. The column variable cv may be a string or just a variable name (which this method will coerce to a string). ## Examples ```r dat <- data.frame(x = gl(4, 25), y = sample(c("A", "B", "C", "D", "E"), 100, replace= TRUE)) pctable(y ~ x, dat) pctable(y ~ x, dat, exclude = NULL) pctable(y ~ x, dat, rvlab = "My Outcome Var", cvlab = "My Columns") pctable(y ~ x, dat, rowpct = TRUE, colpct = FALSE) pctable(y ~ x, dat, rowpct = TRUE, colpct = TRUE) pctable(y ~ x, dat, rowpct = TRUE, colpct = TRUE, exclude = NULL) tab <- pctable(y ~ x, dat, rvlab = "Outcome", cvlab = "Predictor") dat <- data.frame(x1 = gl(4, 25, labels = c("Good", "Bad", "Ugly", "Indiff")), x2 = gl(5, 20, labels = c("Denver", "Cincy", "Baltimore", "NY", "LA")), y = sample(c("A", "B", "C", "D", "E"), 100, replace= TRUE)) tab <- pctable(y ~ x1, data = dat, rvlab = "my row label", subset = dat$x1 %in% c("Good", "Bad"), drop.unused.levels = TRUE) tab <- pctable(y ~ x1, data = dat, rvlab = "my row label", subset = dat$x1 %in% c("Good", "Bad")) pctable("y", "x1", dat) pctable("y", x1, dat) tab <- pctable(y ~ x2, data = dat, rvlab = "A Giant Variable") summary(tab, rowpct = TRUE, colpct = FALSE) tabsum <- summary(tab) ## if user has tables package, can push out to latex or html if (require(tables) & require(Hmisc)){ tabsumtab <- tables::as.tabular(tabsum) Hmisc::html(tabsumtab) fn <- tempfile(pattern = "file", tmpdir = tempdir(), fileext = ".html") Hmisc::html(tabsumtab, file = fn) print(paste("The file saved was named", fn, "go get it.")) if (interactive()) browseURL(fn) unlink(fn) ## go get the fn file if you want to import it in document ## Now LaTeX output ## have to escape the percent signs tabsumtab <- apply(tabsumtab, 1:2, function(x) {gsub("%", "\\\\%", x) }) fn2 <- tempfile(pattern = "file", tmpdir = tempdir(), fileext = ".tex") Hmisc::latex(tabsumtab, file = fn2) print(paste("The file saved was named", fn2, "go get it.")) } ``` ## See Also `tabular` and the CrossTable function in `gmodels` package. ## Author(s) Paul Johnson pauljohn@ku.edu

pctable function