Choice of y-axis can mislead readers

Abstract

Using two examples from the non-scientific literature, we show how choice of unit of measure and scaling of y-axis can caused a biased perception of data, a phenomenon we propose to call perception bias. We recommend to pre-specify unit of measure or how it will be determined, whether outcome variables will be shown as absolute or relative/normalized changes, and to typically start y-axis at 0 for ratio variables.