plotuv: Plot multiple bivariate relationships in a single plot

Description

This function is designed to accompany rm_uvsum as a means of visualising the results, and uses similar syntax.

Usage

plotuv(
  response,
  covs,
  data,
  showN = FALSE,
  showPoints = TRUE,
  na.rm = TRUE,
  response_title = NULL,
  return_plotlist = FALSE,
  ncol = 2,
  p_margins = c(0, 0.2, 1, 0.2),
  bpThreshold = 20,
  mixed = TRUE,
  violin = FALSE,
  position = c("dodge", "stack", "fill"),
  use_labels = TRUE
)

Value

a list containing plots for each variable in covs

Arguments

response: character vector with names of columns to use for response
covs: character vector with names of columns to use for covariates
data: dataframe containing your data
showN: boolean indicating whether sample sizes should be shown on the plots
showPoints: boolean indicating whether individual data points should be shown when n>20 in a category
na.rm: boolean indicating whether na values should be shown or removed
response_title: character value with title of the plot
return_plotlist: boolean indicating that the list of plots should be returned instead of a plot, useful for applying changes to the plot, see details
ncol: the number of columns of plots to be display in the ggarrange call, defaults to 2
p_margins: sets the TRBL margins of the individual plots, defaults to c(0,0.2,1,.2)
bpThreshold: Default is 20, if there are fewer than 20 observations in a category then dotplots, as opposed to boxplots are shown.
mixed: should a mix of dotplots and boxplots be shown based on sample size? If false then all categories will be shown as either dotplots, or boxplots according the bpThreshold and the smallest category size
violin: Show violin plots instead of boxplots. This will override bpThreshold and mixed.
position: for categorical variables how should barplots be presented. Default is "dodge" IF stack is TRUE then n will not be shown.
use_labels: boolean, default is true if the variables have label attributes this will be shown in the plot instead of the variable names, or if there are no labels then tidy versions of the variable names will be used. If use_labels=FALSE the variable names will be used.

Details

Plots are displayed as follows: If response is continuous For a numeric predictor scatterplot For a categorical predictor: If 20+ observations available boxplot, otherwise dotplot with median line If response is a factor For a numeric predictor: If 20+ observations available boxplot, otherwise dotplot with median line For a categorical predictor barplot Response variables are shown on the ordinate (y-axis) and covariates on the abscissa (x-axis)

Variable names are replaced by their labels if available, or by tidy versions if not. Set use_labels=FALSE to use the variable names.

Examples

Run this code

## Run multiple univariate analyses on the pembrolizumab dataset to predict cbr and
## then visualise the relationships.
data("pembrolizumab")
rm_uvsum(data=pembrolizumab,
response='cbr',covs=c('age','sex','l_size','baseline_ctdna'))
plotuv(data=pembrolizumab,  response='cbr',
covs=c('age','sex','l_size','baseline_ctdna'),showN=TRUE)