Computing Effect Sizes for Multi-Level Studies

Software to compute effect sizes for cluster-randomized trials

The main products of a randomized controlled trial include the standardized mean difference (d) and an estimate of its precision. This index is uniquely useful because it is not dependent on the outcome employed in the study. Also, it has the same meaning regardless of whether the study used independent groups, a pre-post design, or a cluster-randomized design (CRT). This allows us to compare the impact of different interventions on the same scale, even if those interventions were tested using different measures or different study designs.

This computer program enables researchers to enter summary data from a cluster-randomized trial and compute the correct estimate of d and its variance. As such, it potentiates the value of these studies in a critically important way. The program is built on an array of sophisticated algorithms, but presents the user with a clear and intuitive interface.

• The program was created to work with all study designs that have been funded by IES. This includes two-level designs and three-level designs, with randomization at any level.
• It also allows researchers to work with any kind of analysis that they may use (including all kinds of analyses that have appeared in published papers).
• It provides clear instructions so the researcher knows what data to enter and how to interpret the output. It also provides extensive documentation for statisticians.

The program was developed with funding from IES, under Dr. Ed Metz. We gratefully acknowledge this funding, but this does not imply an endorsement by IES, which does not endorse any programs as a matter of policy.