Skip to contents

Mixed Correlation Matrix for Meta-Analytic Path Modeling

mixed_corr_meta.Rd

Compute a mixed-type correlation matrix for an effect-size outcome and moderator variables, using inverse-variance weights for pairs involving the outcome and unweighted correlations for moderator-moderator pairs. The function also bootstraps the variance-covariance matrix of the pooled correlations and returns a mars-compatible object for path_model.

Usage

mixed_corr_meta(
  data,
  outcome = "yi",
  variance = "vi",
  moderators,
  variable_types = NULL,
  bootstrap = 1000,
  seed = NULL,
  use_fisher_z = TRUE,
  min_non_missing = 4
)

Arguments

data

A data frame containing the outcome, variance, and moderator columns.

outcome

Name of the effect-size column (for example, "yi").

variance

Name of the sampling-variance column used to define inverse-variance weights (for example, "vi").

moderators

Character vector of moderator variable names.

variable_types

Optional named character vector with values in c("continuous","binary","ordinal"). When NULL, types are inferred from the columns.

bootstrap

Number of bootstrap resamples for estimating the variance-covariance matrix.

seed

Optional random seed for reproducibility.

use_fisher_z

Logical; if TRUE, bootstrap covariance is computed on Fisher-z scale and transformed back to the correlation scale.

min_non_missing

Minimum non-missing pairwise observations required to compute a correlation.

Value

A list with class "mars_corr":

corr_matrix

Pooled correlation matrix.

varcov_matrix

Bootstrap variance-covariance matrix of pooled correlations in corpcor::sm2vec(corr_matrix) order.

method_matrix

Matrix recording the method used for each pair.

pair_details

Data frame with pair, estimate, weighting, and method details.

mars_object

A compact object directly consumable by path_model.

Examples

out <- mixed_corr_meta(
  data = teacher_expectancy,
  outcome = "yi",
  variance = "vi",
  moderators = c("year", "weeks", "setting", "tester"),
  bootstrap = 50,
  seed = 123
)

model <- "yi ~ year + weeks + setting + tester"
fit <- path_model(out, model = model, adjust_se = TRUE)