Skip to contents

Set up a (possibly disconnected) component network for cpaic

Source: R/data_setup.R
cpaic_network.Rd

Builds the network object consumed by cnma_bridge(), cmaic(), cstc() and (Phase 2) cmlnmr(). Aggregate data are supplied at the contrast level (one row per pairwise comparison: treat1 vs treat2, with a treatment effect TE and standard error seTE), matching the input convention of netmeta::discomb(). Individual patient data (IPD) are optional and used only by the population-adjusted methods to replace a study's unadjusted contrast with an adjusted one.

Usage

cpaic_network(
  agd,
  ipd = NULL,
  treat1 = "treat1",
  treat2 = "treat2",
  TE = "TE",
  seTE = "seTE",
  studlab = "studlab",
  sm,
  inactive = NULL,
  sep.comps = "+",
  reference = NULL,
  family = NULL,
  ipd_study = ".study",
  ipd_trt = ".trt",
  ipd_outcome = ".y",
  ipd_time = NULL,
  ipd_status = NULL,
  ipd_exposure = NULL,
  ipd_covariates = NULL
)

Arguments

agd

Aggregate (contrast-level) data frame.

ipd

Optional individual patient data frame (one row per patient). Must contain a study column, a treatment column, an outcome, and the effect-modifier / prognostic covariates.

treat1, treat2, TE, seTE, studlab

Column names in agd.

sm

Summary measure (e.g. "OR", "RR", "MD", "HR"), passed to netmeta::discomb().

inactive

Name of the inactive component / reference (e.g. "placebo").

sep.comps

Component separator in treatment labels. Default "+".

reference

Reference treatment for reported relative effects. Defaults to inactive when available, otherwise the first treatment.

family

Outcome family for the IPD model, one of "binomial", "gaussian", "poisson", "survival" (required when ipd is given).

ipd_study, ipd_trt, ipd_outcome

Column names in ipd. For survival outcomes use ipd_time/ipd_status instead of (or alongside) ipd_outcome; for Poisson rates an optional ipd_exposure offset.

ipd_time, ipd_status

Time and event-indicator column names in ipd (survival family).

ipd_exposure

Optional exposure/person-time column in ipd (Poisson family); used as a log offset.

ipd_covariates

Character vector of covariate column names in ipd (the candidate effect modifiers / prognostic factors).

Value

An object of class cpaic_network: a list with elements agd, ipd, treatments, components, C.matrix, sm, inactive, reference, sep.comps, family, and the column-name mappings.

Details

Treatment labels encode components via sep.comps (e.g. "A + B"), so that sub-networks sharing components can be bridged.

See also

cpaic_connectivity(), cnma_bridge(), cmaic(), cstc()

Examples

# Aggregate-only, disconnected network
net <- cpaic_network(cpaic_bin_agd, sm = "OR", inactive = "Placebo")
net
#> cpaic component network
#>   Summary measure:   OR
#>   Treatments:        6
#>   Components:        4 (A, B, C, D)
#>   AgD comparisons:   5
#>   Reference:         Placebo
#>   Inactive:          Placebo
#>   IPD studies:        none
#>   Connected:         FALSE | components bridgeable: TRUE

# With individual patient data for the population-adjusted methods
net_ipd <- cpaic_network(cpaic_bin_agd, ipd = cpaic_bin_ipd, sm = "OR",
                         family = "binomial", ipd_covariates = "x1",
                         inactive = "Placebo")