Tutorial #3:
Select the structure of the data

A. Declare the structure of the data

By default, the metaumbrella app assumes that all effect sizes are independent and comes from independent studies. If all factors in the umbrella review have only one effect size per study, you can thus skip this part.
In contrast, if, for any given factor, a study included several effect sizes, you have to indicate the presence of such multivariate data using the 3 - Select the structure of your data selector.
If you try to run the analyses on a dataset with multiple effect sizes per study without selecting this option, an error message will be produced. The message will automatically disappear if you rerun the analyses but selecting the "Several effect sizes per study" option.

When multiple outcomes are assessed within the same study, the Borenstein's method to aggregate these dependent effect sizes is used. This method requires to fix a mean correlation coefficient between the outcomes.
You can indicate this information using the r column of your dataset, but you can also fix this value using the slider that appears when indicating that your data have multiple effect sizes per study. The value indicated via the slider will be used to aggregate studies with an empty value in the r column of the dataset.
A mean correlation of .50 is used by default (Wampold et al., 1997) but this value can be adjusted as needed depending on the type of outcomes involved.
When the multiple effect sizes per study are produced by multiple groups, modifying the correlation coefficient will not affect the results.