bivariate_compare {describedata}

Create publication-style table across one categorical variable

Description

Descriptive statistics for categorical variables as well as normally and non-normally distributed continuous variables, split across levels of a categorical variable. Depending on the variable type, an appropriate statistical test is used to assess differences across levels of the comparison variable.

Usage

bivariate_compare(df, compare, normal_vars = NULL,
  non_normal_vars = NULL, cat_vars = NULL, display_round = 2,
  p = TRUE, p_round = 4, include_na = FALSE, col_n = TRUE,
  cont_n = FALSE, all_cont_mean = FALSE, all_cont_median = FALSE,
  iqr = TRUE, fisher = FALSE, workspace = NULL, var_order = NULL,
  var_label_df = NULL)

Arguments

df

A data.frame or tibble.

compare

Discrete variable. Separate statistics will be produced for each level, with statistical tests across levels. Must be quoted.

normal_vars

Character vector of normally distributed continuous variables that will be included in the descriptive table.

non_normal_vars

Character vector of non-normally distributed continuous variables that will be included in the descriptive table.

cat_vars

Character vector of categorical variables that will be included in the descriptive table.

display_round

Number of decimal places displayed values should be rounded to

p

Logical. Should p-values be calculated and displayed? Default TRUE.

p_round

Number of decimal places p-values should be rounded to.

include_na

Logical. Should NA values be included in the table and accompanying statistical tests? Default FALSE.

col_n

Logical. Should the total number of observations be displayed for each column? Default TRUE.

cont_n

Logical. Display sample n for continuous variables in the table. Default FALSE.

all_cont_mean

Logical. Display mean (sd) for all continuous variables. Default FALSE results in mean (sd) for normally distributed variables and median (IQR) for non-normally distributed variables. Must be FALSE if all_cont_median == TRUE.

all_cont_median

Logical. Display median (sd) for all continuous variables. Default FALSE results in mean (sd) for normally distributed variables and median (IQR) for non-normally distributed variables. Must be FALSE if all_cont_mean == TRUE.

iqr

Logical. If the median is displayed for a continuous variable, should interquartile range be displayed as well (TRUE), or should the values for the 25th and 75th percentiles be displayed (FALSE)? Default TRUE

fisher

Logical. Should Fisher's exact test be used for categorical variables? Default FALSE. Ignored if p == FALSE.

workspace

Numeric variable indicating the workspace to be used for Fisher's exact test. If NULL, the default, the default value of 2e5 is used. Ignored if fisher == FALSE.

var_order

Character vector listing the variable names in the order results should be displayed. If NULL, the default, continuous variables are displayed first, followed by categorical variables.

var_label_df

A data.frame or tibble with columns "variable" and "label" that contains display labels for each variable specified in normal_vars, non_normal_vars, and cat_vars.

Details

Statistical differences between normally distributed continuous variables are assessed using aov(), differences in non-normally distributed variables are assessed using kruskal.test(), and differences in categorical variables are assessed using chisq.test() by default, with a user option for fisher.test() instead.

Value

A data.frame with columns label, overall, a column for each level of compare, and p.value. For normal_vars, mean (SD) is displayed, for non_normal_vars median (IQR) is displayed, and for cat_vars n (percent) is displayed. For p values on continuous variables, a superscript 'a' denotes the Kruskal-Wallis test was used

Examples

bivariate_compare(iris, compare = "Species", normal_vars = c("Sepal.Length", "Sepal.Width"))

bivariate_compare(mtcars, compare = "cyl", non_normal_vars = "mpg")