Detecting Careless Responding: A Comparison of Norm Group-Based Mahalanobis Distance and Norm Group-Based Person-Total Correlation with Their Sample-Based Counterparts

Detecting careless responding in self-report questionnaires has gained considerable attention. Different detection methods have been proposed, with implicit methods offering clear advantages over explicit ones. Mahalanobis distance (MD) and person-total correlation ( r _pbis ) are among the most promising implicit detection methods. However, previous research has suggested that MD and r _pbis are less accurate indicators of careless responding when the percentage of careless respondents in the sample is high and that norm group-based (NGB) MD and r _pbis might be more accurate alternatives under that condition. In two simulation studies and a reanalysis of real data, we examined whether NGB MD and NGB r _pbis indeed outperform their sample-based (SB) counterparts in a variety of careless responding patterns. The three main findings were as follows: (1) NGB MD outperformed SB MD in most conditions in three studies, (2) both MD indices (but not r _pbis indices) were ineffective when simulated scale-midpoint invariant and full straightlining careless responding had to be detected, and (3) NGB r _pbis almost never outperformed SB r _pbis . We therefore recommend using NGB MD instead of SB MD, and using SB r _pbis instead of NGB r _pbis when screening data for careless responding.