Transl Clin Pharmacol. 2020 Jun;28(2):83-91. doi: 10.12793/tcp.2020.28.e9. Epub 2020 Jun 24.

Validation of "sasLM," an R package for linear models with type III sum of squares

Jung Sunwoo, Hyungsub Kim, Dohyun Choi, Kyun-Seop Bae

Abstract

The general linear model (GLM) describes the dependent variable as a linear combination of independent variables and an error term. The GLM procedure of SAS® and the "car" package in R calculate the type I, II, or III ANOVA (analysis of variance) tables. In this study, we validated the newly-developed R package, "sasLM," which is compatible with the GLM procedure of SAS®. The "sasLM" package was validated by comparing the output with SAS®, which is the current gold standard for statistical programming. Data from ten books and articles were used for validation. The results of the "sasLM" and "car" packages were compared with those in SAS® using 194 models. All of the results in "sasLM" were identical to those of SAS®, whereas more than 20 models in "car" showed different results from those of SAS®. As the results of the "sasLM" package were similar to those in SAS® PROC GLM, the "sasLM" package could be a viable alternative method for calculating the type II and III sum of squares. The newly-developed "sasLM" package is free and open-source, therefore it can be used to develop other useful packages as well. We hope that the "sasLM" package will enable researchers to conveniently analyze linear models.

Keywords: Analysis of Variance; Bioequivalence; Linear Model; Software Validation; Statistics.

PMID: 32656159  PMCID: PMC7327187  DOI: 10.12793/tcp.2020.28.e9