CODEWITHSUNDEEP
rbayesian
Stats_Dunce
Stats_Dunce

Reputation: 23

Bayesian Network Meta-Analysis (gemtc)

Im trying to conduct a NMA using the R package gemtc. I have three comparisons: A v D: MD: 119 (17.6) B v D: MD: 73 (23.98) C v D: MD: 92 (21.94)

I am following an example code which appear to be functional:

library(gemtc)
library(rjags)
library(dmetar)

data(TherapyFormatsGeMTC)

head(TherapyFormatsGeMTC$data)

TherapyFormatsGeMTC$treat.codes
network <- mtc.network(data.re  = TherapyFormatsGeMTC$data,
                       treatments = TherapyFormatsGeMTC$treat.codes)


model <- mtc.model(network,
                   likelihood = "normal",
                   link = "identity",
                   linearModel = "random",
                   n.chain = 4)

however when i try and use my code:

library(gemtc)
library(rjags)
library(dmetar)

treatments <- read.table(textConnection('
  id  description
  A   "Treatment A"
  B   "Treatment B"
  C   "Treatment C"
  D   "Treatment D"'), header=TRUE)
data <- read.table(textConnection('
  study                 diff         std.err            treatment  
  "Study A"             119          17.60              A          
  "Study A"             NA           NA                 D          
  "Study B"             73           23.98              B          
  "Study B"             NA           NA                 D          
  "Study C"            92           21.94              C          
  "Study C"            NA           NA                 D'), header=TRUE)

data$diff<-as.numeric(data$diff)

network1 <- mtc.network(data, description="Example", treatments=treatments)

model <- mtc.model(network1,
                   likelihood = "normal",
                   link = "identity",
                   linearModel = "random",
                   n.chain = 4)

I get the following error:

Error in validate.data.normal.identity(list(study = c(1L, 1L, 2L, 2L,  : 
  all(data.ab[["std.err"]] > 0) is not TRUE

and I cant work out what I am doing wrong

Upvotes: 0

Views: 628

Answers (1)

程显通
程显通

Reputation: 11

there. The first argument of mtc.network is data.ab, which means data for arms other than relative data, whereas the data in both data mentioned are relative data. Thus, it might be helpful if the data can be passed as data.re rather than data.ab like the following code.

network1 <- mtc.network(data.re = data, description="Example", treatments=treatments)

Upvotes: 1

Related Questions