This R package serves to calculate, via simulation, power and appropriate “stopping” alpha boundaries (and/or futility bounds) for sequential analyses (i.e., “group sequential design”) as well as for multiple hypotheses (multiple tests included in an analysis), given any specified global error rate. This enables the sequential use of practically any significance test, as long as the underlying data can be simulated in advance to a reasonable approximation.
For how to use, see the vignettes linked below.
To install the stable version from CRAN, just run:
install.packages("POSSA")Then load with: library("POSSA")
Alternatively, if you want to install the latest (and potentially unstable) version from this repository:
install.packages("devtools") # if "devtools" package is not yet installed
library("devtools")
install_github("gasparl/possa")Please see the vignettes:
(Less important: Benchmarking)
For detailed information about each function and parameter, see the manual.
Given that this project is not part of my “day job”, the intended additional features grew beyond my capacity. Help would be most welcome; if you are up to it, feel free to email me (or to reply to the relevant issue or open a new one; see details here).
The name “POSSA” refers not to the Italian meaning “power, strength” (/ˈpɔs.sa/), but to the plural of “possum” (/ˈpɑsə/).
When you use POSSA in a publication, you can either cite the specific
version you used (enter citation("POSSA") in R), or the
following paper:
Lukács, G. (2022). POSSA: Power simulation for sequential analyses and multiple hypotheses. Journal of Open Source Software, 7(76), 4643. https://doi.org/10.21105/joss.04643