eReport {greport}

Event Report

Description

Generates graphics for binary event proportions

Usage

eReport(
  formula,
  data = NULL,
  subset = NULL,
  na.action = na.retain,
  minincidence = 0,
  conf.int = 0.95,
  etype = "adverse events",
  panel = "events",
  subpanel = NULL,
  head = NULL,
  tail = NULL,
  h = 6,
  w = 7,
  append = FALSE,
  popts = NULL
)

Arguments

formula

a formula with one or two left hand variables (the first representing major categorization and the second minor), and 1-2 right hand variables. One of these may be enclosed in id() to indicate the presence of a unique subject ID, and the other is treatment.

data

input data frame

subset

subsetting criteria

na.action

function for handling NAs when creating analysis frame

minincidence

a number between 0 and 1 specifying the minimum incidence in any stratum that must hold before an event is included in the summary

conf.int

confidence level for difference in proportions

etype

a character string describing the nature of the events, for example "adverse events", "serious adverse events". Used in figure captions.

panel

panel string

subpanel

a subpanel designation to add to panel

head

character string. Specifies initial text in the figure caption, otherwise a default is used.

tail

a character string to add to end of automatic caption

h

height of graph

w

width of graph

append

set to TRUE if adding to an existing sub-report

popts

a list of options to pass to graphing functions

Details

Generates dot charts showing proportions on left and risk difference with confidence intervals on the right, if there is only one level of event categorization. Input data must contain one record per event, with this record containing the event name. If there is more than one event of a given type per subject, unique subject ID must be provided. Denominators come from greport options and it is assumed that only randomized subjects have records. Some of the graphics functions are modifications of those found in the HH package. The data are expected to have one record per event, and non-events are inferred from setgreportOption('denom'). It is also assumed that only randomized subjects are included in the dataset.

Author(s)

Frank Harrell

Examples

# See test.Rnw in tests directory