search

tklebel / crosstabr

Cross tabulation with HTML and Markdown output
Other
1 stars 0 forks source link

readme

Travis-CI Build Status AppVeyor Build Status Coverage Status CRAN_Status_Badge lifecycle

crosstabr

crosstabr is a package to create crosstabs with a simple syntax. Current implementations like gmodels::CrossTable lack a simple and understandable way of setting up all common options and deliver unusable output if there are many columns.

The default output for crosstab is HTML, similar to ggvis. The resulting crosstable can be viewed within RStudio's viewer pane, or can be displayed in a separate browser window.

Currently crosstab lacks compatibility with RMarkdown, which will be added in the future.

Installation

# install.packages("devtools")
devtools::install_github("tklebel/crosstabr")

Basic Usage

The package builds upon the idea, that when investigating categorial data, in most cases there will be some sort of dependent variable: a variable the researcher is interested. On the other hand there will be one or more independent variables, where the goal is to determine, if there is a difference in the DV regarding to different outcomes of the IV's. Usually one would calculate percentages on rows or columns, depending on if the DV is in the rows or cols.

crosstabrs syntax is similar to linear modeling: you specify a model via a formula, with the DV on the left-hand-side, the IV on the right-hand-side. For example, we could be interested whether females or males were more likely to survive the sinking of the Titanic. We would therefore place the DV (Survived) on the left-hand-side, the IV (Sex) on the right-hand-side:

titanic %>% 
  crosstab(Survived ~ Sex)

If no layout is provided crosstabr defaults to layout_column, where the IV is placed in the columns. We could however be explicit about the layout:

titanic %>% 
  crosstab(Survived ~ Sex) %>% 
  layout_column()

Currently layout_row() is not yet implemented, but it is planned for a future release.

Statistical Tests

In many cases a simple table won't be enough; possibly one wants to conduct statistical tests on the table. This is fairly easy:

titanic %>% 
  crosstab(Survived ~ Sex) %>% 
  add_stats()

When no specific tests are supplied, add_stats() defaults to vcd::assocstats() which yields the result of a chisq.test, the likelihood ratio and three coefficients:

If desired, other tests can be supplied:

titanic %>% 
  crosstab(Survived ~ Sex) %>% 
  add_stats(fisher.test)

It is possible to supply several functions, as well as anonymous functions in order to add arguments:

titanic %>% 
  crosstab(Survived ~ Sex) %>% 
  add_stats(chisq.test, function(x) fisher.test(x, alternative = "greater"))