The CRTspat
package is intended to support the design
and analysis of cluster-randomized trials (CRTs) where there is
geographical contamination between the arms. It can also be used in
simulation of such CRTs to support development of improved methodology.
An important set of applications is in field trials of vector-control
interventions against malaria.
The package can facilitate design of efficient trials by providing algorithms for:
The package builds on the work of Multerer et al. (2021a), Multerer et al. (2021b) and Anaya-Izquierdo & Alexander(2021).
The package is intended to function with outcomes that are proportions, count data, or continuous variables. The current version (0.1.0.9000) is for testing and has been developed mainly on data that are proportions. It will be revised to ensure that it gives the correct outputs with other kinds of outcome.
Examples are provided of the workflow for different use cases, as follows:
1. Algorithmic definition of clusters
2. Simulation of trials with geographical contamination
3. Estimation of intracluster correlations (ICC) by cluster size
4. Estimation of optimal cluster size for a trial with pre-determined buffer width
5. Analysis of trials (including methods for analysing contamination)
6. Thematic mapping of the geography of a CRT
7. Power and sample size calculations allowing for contamination
8, Eggs-to fry or scramble? Implications of excluding buffer zones from the measurement of outcomes
9: Preparation of datasets for
CRTspat
10: Anonymising input locations (if required for sharing reference datasets)
The package functions are listed in the manual.