Skip to contents

Customising the models to compare

Source: vignettes/customise_models.Rmd
customise_models.Rmd
library(metaUI)
#> Note re metaUI: This package is still under active development - so please report any issues and desired features on https://github.com/LukasWallrich/metaUI.

Adding, editing and removing models

You might wish to show different models. For now, the easiest way to do that is to save the app into a folder, as shown in the vignette(“getting_started”). Then open models.R to edit the models.

Removing models

To remove models, you can add a comment (#) in front of the line where it is first defined in upper part of the file where models_to_run are defined … and ignore the longer part in the models_code below. So to remove the Trim-and-fill model, you simply need to add the # shown below - or alternatively to delete that line.

   "Robust Variance Estimation",                      FALSE,       "as.numeric(mod$reg_table$b.r)",      ... 
#   "Trim-and-fill",                                   TRUE,        "mod$TE.random",                     ...
   "P-uniform star",                                  TRUE,        "mod$est",                            ...

Add or edit models

To add or edit models, you need to understand a bit more about the structure of the file. If contains two parts - firstly, the models_to_run which has the names of the models and the details on how to extract summary data, and then the models_code that contains the code to estimate them.

For models_to_run, you need to provide the following details for each model:

  • name of the model to be displayed
  • aggregated TRUE/FALSE value whether it needs to be estimated with the dataset aggregated by sample (i.e. without dependent effect sizes)

And then code within “” that allows you to export the following from the model object, called mod:

  • es: the effect size
  • LCL & UCL: the lower and upper bounds of the 95% confidence intervals for the effect size
  • k: the number of effects that the model was estimated on

For models_code, you need to provide code, again in ““, that estimates the model based on fields available in the dataset produced by prepare_data() (for models where aggregated = FALSE) or produced as df_agg by server.R By default, this only includes metaUI__study_id, metaUI__effect_size, metaUI__variance and metaUI__se, so if you need any other fields here, you need to adjust server.R as well.

If your model code is complicated, it might be worthwhile to create helper functions to call from within the models_code entry. You can add them to models.R but they need to be assigned to the global environment using <<-, e.g., my_meta_fun <<- function(...)