tbl_propdiff: Calculating unadjusted and adjusted differences in rates

In ddsjoberg/hotfun: Collection of Functions Used in the Health Outcomes Team at MSKCC

Description

This function calculates the unadjusted or adjusted difference in rates with confidence interval.

Usage

tbl_propdiff(
  data,
  y,
  x,
  formula = "{y} ~ {x}",
  label = NULL,
  statistic = "{n} ({p}%)",
  method = c("chisq", "exact", "boot_sd", "boot_centile"),
  conf.level = 0.95,
  bootstrapn = ifelse(method == "boot_centile", 2000, 200),
  estimate_fun = style_sigfig,
  pvalue_fun = style_pvalue
)

Arguments

data	A data frame
y	vector of binary outcome variables. Outcome variables can be numeric, character or factor, but must have two and only two non-missing levels
x	string indicating the binary stratifying variable. The stratifying variable can be numeric, character or factor, but must have two and only two non-missing levels
formula	By default, "{y} ~ {x}". To include covariates for an adjusted risk difference, add covariate names to the formula, e.g. "{y} ~ {x} + age"
label	List of formulas specifying variables labels, If a variable's label is not specified here, the label attribute (attr(data$high_grade, "label")) is used. If attribute label is NULL, the variable name will be used.
statistic	Statistics to display for each group. Default "{n} ({p}%)"
method	The method for calculating p-values and confidence intervals around the difference in rates. The options are "chisq", "exact", "boot_centile", and "boot_sd". See below for details. Default method is "chisq".
conf.level	Confidence level of the returned confidence interval. Must be a single number between 0 and 1. The default is a 95% confidence interval.
bootstrapn	The number of bootstrap resamples to use. The default is 2000 for "boot_centile" and 200 for "boot_sd"
estimate_fun	Function to round and format estimates. By default style_sigfig, but can take any formatting function
pvalue_fun	Function to round and format p-value. By default style_pvalue, but can take any formatting function

Value

A tbl_propdiff object, with sub-class "gtsummary"

Methods

• The chisq option returns a p-value from the prop.test function and a confidence interval for the unadjusted difference in proportions based on the normal approximation.

• The exact option returns a p-value from the fisher.test function. The confidence interval returned by this option is the same as the confidence interval returned by the chisq option and is based on the normal approximation.

• The boot_centile option calculates the adjusted difference between groups in all bootstrap samples (the default for this method is 2000 resamples) and generates the confidence intervals from the distribution of these differences. For the default, a 95% confidence interval, the 2.5 and 97.5 centiles are used. The p-value presented is from a logistic regression model.

• The boot_sd option calculates the adjusted difference between groups in all bootstrap samples (the default for this method is 200 resamples). The mean and standard deviation of the adjusted difference across all resamples are calculated. The standard deviation is then used as the standard error to calculate the confidence interval based on the true adjusted difference. The p-value presented is from a logistic regression model.

Examples

tbl_propdiff(
  data = trial,
  y = "response",
  x = "trt"
)

tbl_propdiff(
  data = trial,
  y = "response",
  x = "trt",
  formula = "{y} ~ {x} + age + stage",
  method = "boot_sd",
  bootstrapn = 25
)

ddsjoberg/hotfun documentation built on Dec. 12, 2021, 3:41 a.m.