On coordinating “simple” statistical analyses across multiple software packages: A case study from wave 1 of the Global Flourishing Study

Statistical analyses can have a significant impact when the results are easy to communicate to a wide audience. This goal was a focus when designing the coordinated analyses of Wave 1 of the Global Flourishing Study. The methods, especially for the demographic variation analyses, are seemingly simple: to compute the mean or proportion of an outcome by subgroups. When described at a high level, these analyses feel like a topic that should be only of interest while in an introductory statistics course. The current work will challenge this belief as we describe our efforts to provide a rigorous methodology for these seemingly simple analyses that can be implemented across a range of popular statistical software packages. The descriptions of the methods are supplemented by a focused simulation study demonstrating the consistency of results of complex survey adjusted modified Poisson regression with missing data across statistical software packages R, Stata, and SAS.