netmeta (version 3.1-1)

rankogram.default: Calculate rankogram from treatment effect samples

Description

This function calculates the probabilities of each treatment being at each possible rank and the SUCRAs (Surface Under the Cumulative RAnking curve) from a sample of treatment estimates in network meta-analysis.

Usage

# S3 method for default
rankogram(
  x,
  pooled = "unspecified",
  small.values = "desirable",
  cumulative.rankprob = FALSE,
  keep.samples = FALSE,
  nchar.trts = gs("nchar.trts"),
  ...
)

Value

An object of class rankogram with corresponding print

and plot function. The object is a list containing the following components:

ranking.matrix.common

Numeric matrix giving the probability of each treatment being at each possible rank for the common effects model.

ranking.common

SUCRA values for the common effects model.

ranking.matrix.random

Numeric matrix giving the probability of each treatment being at each possible rank for the random effects model.

ranking.random

SUCRA values for the random effects model.

cumrank.matrix.common

Numeric matrix giving the cumulative ranking probability of each treatment for the common effects model.

cumrank.matrix.random

Numeric matrix giving the cumulative ranking probability of each treatment for the random effects model.

nsim, common, random

As defined above

,

small.values, x

As defined above

,

Arguments

x

A matrix or data frame with treatment effects in columns and samples in rows.

pooled

A character string indicating whether samples come from a common ("common"), random effects ("random"), or "unspecified" model, can be abbreviated.

small.values

An optional character string specifying whether small treatment effects indicate a beneficial ("desirable") or harmful ("undesirable") effect, can be abbreviated.

cumulative.rankprob

A logical indicating whether cumulative ranking probabilities should be printed.

keep.samples

A logical indicating whether to keep the generated samples.

nchar.trts

A numeric defining the minimum number of characters used to create unique treatment names.

...

Additional arguments (ignored).

Author

Theodoros Papakonstantinou , Guido Schwarzer

Details

We derive a matrix showing the probability of each treatment being at each possible rank. To this aim, we use samples and summarise them using the ranking metric SUCRAs (Surface Under the Cumulative RAnking curve).

The matrix / data frame in argument x must contain the sampled effects for each treatment.

References

Salanti G, Ades AE, Ioannidis JP (2011): Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial. Journal of Clinical Epidemiology, 64, 163--71

See Also

netmeta, netrank, plot.rankogram, dat.woods2010, dat.linde2015

Examples

pw1 <- pairwise(treatment, event = r, n = N, studlab = author,
  data = dat.woods2010, sm = "OR")
net1 <- netmeta(pw1, small.values = "desirable")

set.seed(1909) # get reproducible results
ran1 <- rankogram(net1, common = FALSE, nsim = 10, # reduce runtime
  keep.samples = TRUE)
ran1

rankogram(ran1$samples.random, pooled = "random")

# \donttest{
pw2 <- pairwise(treat = list(treatment1, treatment2, treatment3),
  event = list(resp1, resp2, resp3), n = list(n1, n2, n3),
  studlab = id, data = dat.linde2015, sm = "OR")
#
net2 <- netmeta(pw2, common = FALSE,
  ref = "Placebo", small = "undesirable")

ran2 <- rankogram(net2, nsim = 100, common = FALSE,
  keep.samples = TRUE)
ran2

# Wrong ranking due to using the default,
# i.e., argument 'small.values = "desirable".
rankogram(ran2$samples.random, pooled = "random")
# Correct ranking 
rankogram(ran2$samples.random, pooled = "random",
  small.values = "undesirable")
# }