Calculate sample size for and compare adaptive interventions within a clustered SMART with a continuous, end-of-study outcome

About This Software Bundle

A clustered SMART is a type of randomized trial where intact clusters of units (e.g., patients) are randomized sequentially, yet the primary outcome is at the level of the units within the cluster. This software bundle does two things: (i) It provides data analysis code for comparing the embedded adaptive interventions in a clustered SMART on an end of study, continuous outcome. (ii) It provides the code for calculating the minimum sample size necessary for a clustered SMART in which the primary aim is the comparison of two adaptive interventions beginning with different interventions.

How can a behavioral scientist use this software bundle?

Behavioral intervention scientists can use this code to compare the mean of a continuous outcome between the adaptive interventions embedded in a clustered SMART.

The code also can be used to calculate the minimum sample size for clustered SMARTs in which the primary aim is the comparison of two adaptive interventions.

What method does this software bundle implement?

This code implements sample size formulae based on a test of the null hypothesis that the difference in mean between two adaptive interventions in a clustered SMART is zero. The sample size formula is a product of: the sample size formula for a standard two-arm clustered randomized trial, a deflation factor that accounts for the increased statistical efficiency resulting from adjusting for a baseline covariate, and a SMART inflation factor that depends on the non-response rate.