Skip to contents

Meta-analyses multiple datasets or subsets of datasets of analyst data

Source: R/meta_analyse_datasets.R
meta_analyse_datasets.Rd

Runs all meta-analyses and regression models for the ManyEcoEvo project analysis, including:

  • Fitting univariate / fixed-effects meta-analysis

  • Calculating the deviation of every effect size / point-estimate from the meta-analytic mean for all data subsetes

  • The absolute, box-cox transformed deviation scores

  • A univariate GLM regression of the transformed deviation scores on the sorensen diversity indices

  • A univariate GLM regression of the transformed deviation scores on the continuous peer-review ratings

  • A univariate GLM regression of the transformed deviation scores on the categorical peer-review ratings

  • A univariate GLM regression of the transformed deviation scores on a binary variable corresponding to whether the analysis was a mixed-effects model (i.e. GLM with random-effects) or not.

  • To be implemented: a multivariate regression #TODO

  • The deviation scores on transformed categorical ratings but with no intercept (for nice plotting / exploration).

Usage

meta_analyse_datasets(
  data,
  outcome_variable = NULL,
  outcome_SE = NULL,
  filter_vars = NULL
)

Arguments

data

A nested- dataframe grouped by dataset and / or exclusion_set, estimate_type, containing the list-column of prepared analyst subset data effects_analysis ready for meta-analysis.

filter_vars

A list of expressions to filter the data dataframe by. E.g. rlang::exprs(exclusion_set == "complete", expertise_subset == "All", publishable_subset == "All", collinearity_subset == "All") #' @param outcome_variable A named list containing either/and a list of datasets and their corresponding outcome variables for each value of dataset, a list of estimate_types and their corresponding outcome variables for each value of estimate_type.

Value

A nested dataframe with all columns of object parsed to arg data, but with additional columns for the results of each analysis: MA_mod, sorensen_glm, box_cox_ratings_cont, box_cox_ratings_cat, box_cox_rating_cat_no_int, uni_mixed_effects

Details

When filter_vars are supplied the function will filter the data dataframe by the expressions in the list, any data subsets excluded by filtering will not have multivariate met-analysis models fitted with fit_multivar_MA().

When the arguments outcome_variable and/or outcome_variable are not supplied, the function defaults to:

  • using "Zr" as the standardised effect size and "VZr" as the standard error when estimate_type is "Zr".

  • using "Z" as the standardised out-of-sample estimate and "VZ" as the standardised out-of-sample estimate error when estimate_type is one of c("yi", "y25", "y50", "y75").

The function will check if the data dataframe contains the required columns for meta-analysis, including any variable names specified in calls to the filter_vars argument. If the required columns do not exist then the function will stop with an error.

Function assumes that if argument outcome_variable is supplied, then outcome_SE is also supplied, and conversely, if outcome_SE is not supplied, then neither is outcome_variable (TODO not yet checked in function).

See also

Other Multi-dataset Wrapper Functions: apply_VZ_exclusions(), compute_MA_inputs(), generate_exclusion_subsets(), generate_outlier_subsets(), generate_rating_subsets(), generate_yi_subsets(), make_viz(), prepare_ManyEcoEvo(), prepare_ManyEcoEvo_yi(), prepare_response_variables(), prepare_response_variables_yi(), summarise_analysis_types(), summarise_conclusions(), summarise_model_composition(), summarise_reviews(), summarise_sorensen_index(), summarise_variable_counts()

Examples

filter_vars <- rlang::exprs(
  exclusion_set == "complete",
  expertise_subset == "All",
  publishable_subset == "All",
  collinearity_subset == "All"
)