Rank probabilities indicate the probability for each treatment to be best, second best, etc.
rank.probability(result, preferredDirection=1, covariate=NA)# S3 method for mtc.rank.probability
print(x, ...)
# S3 method for mtc.rank.probability
plot(x, ...)
sucra(ranks)
rank.quantiles(ranks, probs=c("2.5%"=0.025, "50%"=0.5, "97.5%"=0.975))
rank.probability: A matrix (with class mtc.rank.probability) with the treatments as rows and the ranks as columns.
sucra: A vector of SUCRA values.
rank.quantiles: A matrix with treatments as rows and quantiles as columns, giving the quantile ranks (by default, the median and 2.5% and 97.5% ranks).
Object of S3 class mtc.result to be used in creation of the rank probability table
Preferential direction of the outcome. Set 1 if higher values are preferred, -1 if lower values are preferred.
(Regression analyses only) Value of the covariate at which to compute rank probabilities.
An object of S3 class rank.probability.
Additional arguments.
A ranking matrix where the treatments are the rows (e.g. the result of rank.probability).
Probabilities at which to give quantiles.
Gert van Valkenhoef, Joël Kuiper
For each MCMC iteration, the treatments are ranked by their effect relative to an arbitrary baseline. A frequency table is constructed from these rankings and normalized by the number of iterations to give the rank probabilities.
relative.effect
model <- mtc.model(smoking)
# To save computation time we load the samples instead of running the model
if (FALSE) results <- mtc.run(model)
results <- readRDS(system.file("extdata/luades-smoking-samples.rds", package="gemtc"))
ranks <- rank.probability(results)
print(ranks)
## Rank probability; preferred direction = 1
## [,1] [,2] [,3] [,4]
## A 0.000000 0.003000 0.105125 0.891875
## B 0.057875 0.175875 0.661500 0.104750
## C 0.228250 0.600500 0.170875 0.000375
## D 0.713875 0.220625 0.062500 0.003000
print(sucra(ranks))
## A B C D
## 0.03670833 0.39591667 0.68562500 0.88175000
print(rank.quantiles(ranks))
## 2.5% 50% 97.5%
## A 3 4 4
## B 1 3 4
## C 1 2 3
## D 1 1 3
plot(ranks) # plot a cumulative rank plot
plot(ranks, beside=TRUE) # plot a 'rankogram'
Run the code above in your browser using DataLab