multi: Multinomial outcome data

In multinma: Bayesian Network Meta-Analysis of Individual and Aggregate Data

Multinomial outcome data

Description

This function aids the specification of multinomial outcome data when setting up a network with set_agd_arm() or set_ipd(). It takes a set of columns (or, more generally, numeric vectors of the same length) of outcome counts in each category, and binds these together to produce a matrix.

Usage

multi(..., inclusive = FALSE, type = c("ordered", "competing"))

Arguments

...	Two or more numeric columns (or vectors) of category counts. Argument names (optional) will be used to label the categories.
inclusive	Logical, are ordered category counts inclusive (TRUE) or exclusive (FALSE)? Default FALSE. Only used when ordered = TRUE. See details.
type	String, indicating whether categories are "ordered" or "competing". Currently only ordered categorical outcomes are supported by the modelling functions in this package.

Details

When specifying ordered categorical counts, these can either be given as exclusive counts (inclusive = FALSE, the default) where individuals are only counted in the highest category they achieve, or inclusive counts (inclusive = TRUE) where individuals are counted in every category up to and including the highest category achieved. (Competing outcomes, by nature, are always specified as exclusive counts.)

NA values can be used to indicate categories/cutpoints that were not measured.

Value

A matrix of (exclusive) category counts

Examples

# These two data sets specify the same ordered categorical data for outcomes
# r0 < r1 < r2, but the first uses the "inclusive" format and the second the
# "exclusive" format.
df_inclusive <- tibble::tribble(~r0, ~r1, ~r2,
                                1, 1, 1,
                                5, 4, 1,
                                5, 2, 2,
                                10, 5, 0,
                                5, 5, 0,
                                7, NA, 6,   # Achieved r2 or not (no r1)
                                10, 4, NA)  # Achieved r1 or not (no r2)

df_exclusive <- tibble::tribble(~r0, ~r1, ~r2,
                                0, 0, 1,
                                1, 3, 1,
                                3, 0, 2,
                                5, 5, 0,
                                0, 5, 0,
                                1, NA, 6,   # Achieved r2 or not (no r1)
                                6, 4, NA)   # Achieved r1 or not (no r2)

(r_inclusive <- with(df_inclusive, multi(r0, r1, r2, inclusive = TRUE)))
(r_exclusive <- with(df_exclusive, multi(r0, r1, r2, inclusive = FALSE)))

# Counts are always stored in exclusive format
stopifnot(isTRUE(all.equal(r_inclusive, r_exclusive)))


## HTA Plaque Psoriasis
library(dplyr)

# Ordered outcomes here are given as "exclusive" counts
head(hta_psoriasis)

# Calculate lowest category count (failure to achieve PASI 50)
pso_dat <- hta_psoriasis %>%
  mutate(`PASI<50` = sample_size - rowSums(cbind(PASI50, PASI75, PASI90), na.rm = TRUE))

# Set up network
pso_net <- set_agd_arm(pso_dat,
                       study = paste(studyc, year),
                       trt = trtc,
                       r = multi(`PASI<50`, PASI50, PASI75, PASI90,
                                 inclusive = FALSE,
                                 type = "ordered"))

pso_net

multinma documentation built on May 31, 2023, 5:46 p.m.