Comparative analysis of regression output summary statistics in common statistical packages

Abstract

An important question in applied statistics is whether statistical regression programs output distorted summary statistics (F and R2) for a no-intercept regression model. Uyar and Erdem recently confirmed this problem using one of the premier statistical packages. This article extends the Uyar-Erdem experiment to eight common statistical and econometric regression packages. Only four of them warn users that the R2 has been redefined for a model without an explicit intercept. Consequently, the problem Uyar and Erdem identified extends in varying degrees, ranging from mild to severe, to many other regression packages. © 1993 Taylor & Francis Group, LLC.

Publication Title

American Statistician

Link to Full Text

Share

COinS