Fitting a network meta-analysis model using INLA

nma_inla fits a network meta-analysis model using INLA.

nma_inla(datINLA, likelihood = NULL, fixed.par = c(0, 1000),
  tau.prior = "uniform", tau.par = c(0, 5), kappa.prior = "uniform",
  kappa.par = c(0, 5), mreg = FALSE, type = "consistency",
  verbose = FALSE, inla.strategy = "simplified.laplace",
  improve.hyperpar.dz = 0.75, correct = FALSE, correct.factor = 10,
  improve.hyperpar = TRUE)

Arguments

datINLA	An object of create_INLA_dat
likelihood	The likelihood to be used.
fixed.par	A numerical vector specifying the parameter of the normal prior density for basic parameters, first value is parameter for mean, second is for variance.
tau.prior	A string specifying the prior density for the heterogeneity standard deviation, options are 'uniform' for uniform prior and 'half-normal' for half-normal prior.
tau.par	A numerical vector specifying the parameter of the prior density for heterogenety stdev. var.par = c(u, l): u is lower bound and l is upper bound when var.prior = 'uniform'. var.par = c(m, v): m is mean and v is variance when var.prior = 'half-normal'.
kappa.prior	A string specifying the prior density for the inconsistency standard deviation, options are 'uniform' for uniform prior and 'half-normal' for half-normal prior.
kappa.par	A numerical vector specifying the parameter of the prior. density for inconsistency stdev. The definition of the priors is the same as for tau.par.
mreg	Logical indicating whether covariate(s) should be incorporated to fit a network meta-regression model, default FALSE.
type	A string indicating the type of the model, options are "FE", "consistency" and "jackson".
verbose	Logical indicating whether the program should run in a verbose model, default FALSE.
inla.strategy	A string specfying the strategy to use for the approximations of INLA; one of 'gaussian', 'simplified.laplace' (default) or 'laplace', see ?INLA::control.inla.
improve.hyperpar.dz	Step length in the standardized scale used in the construction of the grid, default 0.75, see INLA::inla.hyperpar. Not used if mod = 'FE'.
correct	Logical Add correction for the Laplace approximation, default FALSE, see INLA::inla.hyperpar. Not used if mod = 'FE'.
correct.factor	Numerical Factor used in adjusting the correction factor if correct=TRUE, default 10, see INLA::inla.hyperpar. Not used if mod = 'FE'.
improve.hyperpar	Improve the estimates of the posterior marginals for the hyperparameters of the model using the grid integration strategy, default TRUE. see INLA::inla.hyperpar.

Value

nma_inla returns a nma_inla object.

Details

The following likelihood types are supported

• normal: for continuous (mean difference) data.

  Required coloumns: [mean, std.err]

  Result: relative mean difference

• binomial: for dichotomous data.

  Required coloumns: [responders, sampleSize]

  Result: log odds ratio

• normal: for event-rate (survival) data.

  Required coloumns: [responders, exposure]

  Result: log hazard ratio

The following model types are supported

• FE, ordinary fixed effect model, assuming homogeneity between trials (Dias et al., 2013)

• consistency, ordinary consistency model, assuming consistency in the network. (Jackson et al., 2014)

• jackson, the design-by-treatment interaction model with random inconsistency parameters. (Jackson et al., 2014)

Examples

SmokdatINLA <- create_INLA_dat(dat = Smokdat, armVars = c('treatment' = 't', 'responders' = 'r'
,'sampleSize' = 'n'), nArmsVar = 'na')
# NOT RUN {
## Fitting a consistency model
if(requireNamespace('INLA', quietly = TRUE)){
 require('INLA', quietly = TRUE)
fit.Smok.cons.INLA <- nma_inla(SmokdatINLA, likelihood = 'binomial', type = 'consistency',
tau.prior = 'uniform', tau.par = c(0, 5))
}
# }