Fit random-effects Bayesian meta-analytic CFAs with minor factors assumed.

A function to fit fixed- or random-effects Bayesian meta-analytic CFAs with minor factors assumed (Uanhoro 2023) . Builds off work by Wu and Browne (2015) . Does not yet accomodate moderators, and covariance matrices must be complete. This will change in the near future.

Usage

meta_mbcfa(
  model = NULL,
  data = NULL,
  group = NULL,
  sample_cov = NULL,
  sample_nobs = NULL,
  method = "normal",
  type = "re",
  orthogonal = FALSE,
  simple_struc = TRUE,
  seed = 12345,
  warmup = 1000,
  sampling = 1000,
  refresh = (warmup + sampling)/10,
  adapt_delta = 0.9,
  max_treedepth = 10,
  chains = 4,
  ncores = max(parallel::detectCores() - 2, 1),
  priors = new_mbsempriors(),
  show = TRUE,
  show_messages = TRUE,
  cluster = NULL,
  target = "rstan"
)

Arguments

model: A description of the user-specified model, lavaan syntax.
data: An optional data frame containing the observed variables used in the model.
group: An optional string identifying the grouping variable in the data object.
sample_cov: (list of matrices) sample variance-covariance matrices. The rownames and/or colnames must contain the observed variable names. For now, assumes there are no missing elements in the covariance matrices.
sample_nobs: (vector of positive integer) Number of observations for each study.
method: (character) One of "normal", "lasso", "logistic", "GDP", or "none". See details below.
type: (character) One of "fe", "re", or "dep" for fixed-effects, random-effects, and dependent-samples MASEM respectively. The "dep" argument is experimental, see details below.
orthogonal: (logical) constrain factors orthogonal, must be TRUE to fit bifactor models.
simple_struc: (LOGICAL) Only relevant for CFAs. If TRUE: assume simple structure; If FALSE: estimate all cross-loadings using generalized double Pareto priors.
seed: (positive integer) seed, set to obtain replicable results.
warmup: (positive integer) The number of warmup iterations to run per chain.
sampling: (positive integer) The number of post-warmup iterations to run per chain, retained for inference.
refresh: (positive integer) How often to print the status of the sampler.
adapt_delta: (real in (0, 1)) Increase to resolve divergent transitions.
max_treedepth: (positive integer) Increase to resolve problems with maximum tree depth.
chains: (positive integer) The number of Markov chains to run.
ncores: (positive integer) The number of chains to run in parallel.
priors: An object of mbsempriors-class. See new_mbsempriors for more information.
show: (Logical) If TRUE, show table of results, if FALSE, do not show table of results. As an example, use FALSE for simulation studies.
show_messages: (Logical) If TRUE, show messages from Stan sampler, if FALSE, hide messages.
cluster: An optional integer vector identifying the cluster each group belongs to. Asssume there are five groups, the first three belong to cluster 1 and the last two belong to cluster 2, then the argument would be: cluster = c(1, 1, 1, 2, 2). This feature is experimental, see details below.
target: (character) One of "rstan" or "cmdstan". If "cmdstan", CmdStan and CmdStanR need to be installed on the device.

Value

An object of mbsem-class

Details

CFAs assume standardized factors. Latent variable regression models are not yet implemented.

There are different methods for estimating models in this package:

normal: under belief that minor factor influences are on average zero with continuous deviations away from zero.
lasso: under belief that minor factor influences are largely zero with a small number of non-zero residual covariances.
logistic: for similar belief as normal but more readily accomodates extreme outliers.
GDP: to mimic a global-local approach, i.e. attempt to shrink near 0 residual covariances to 0 with minimal shrinking for larger residual covariances (Armagan et al. 2013) .
none: if intending to ignore the influence of minor factors.

When type = "dep", the user must supply the cluster IDs, see cluster parameter documentation above. However, this feature is experimental and only available when target = "cmdstan". Additionally, the cluster inputs are not validated.

References

Armagan A, Dunson DB, Lee J (2013). “Generalized double Pareto shrinkage.” Statistica Sinica, 23(1), 119--143. ISSN 1017-0405, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3903426/.

Uanhoro JO (2023). “Hierarchical Covariance Estimation Approach to Meta-Analytic Structural Equation Modeling.” Structural Equation Modeling: A Multidisciplinary Journal, 30(4), 532--546. ISSN 1070-5511, doi:10.1080/10705511.2022.2142128 , 2023-06-19.

Wu H, Browne MW (2015). “Quantifying Adventitious Error in a Covariance Structure as a Random Effect.” Psychometrika, 80(3), 571--600. ISSN 1860-0980, doi:10.1007/s11336-015-9451-3 .

Examples

if (FALSE) {
model_syntax <- "# latent variable definitions
F1 =~ JP1 + JP2 + JP3
F2 =~ JN1 + JN2 + JN4 + JN4
F3 =~ TD1 + TD2"
meta_mbcfa(
  model_syntax,
  sample_cov = issp89$data, sample_nobs = issp89$n
)
model_syntax <- paste0(
  "distress =~ ", paste0("x", 1:14, collapse = " + "), "\n",
  "anxiety =~ ", paste0("x", seq(1, 14, 2), collapse = " + "), "\n",
  "depression =~ ", paste0("x", seq(2, 14, 2), collapse = " + ")
)
meta_mbcfa(
  model_syntax,
  sample_cov = Norton13$data, sample_nobs = Norton13$n, orthogonal = TRUE
)
}