Missing data bias: Exactly how bad is pairwise deletion?

Daniel A. Newman, Jonathan M. Cottrell

Research output: Chapter in Book/Report/Conference proceedingChapter


When conducting social science research, missing data are a ubiquitous practical difficulty (Anseel, Lievens, Scollaert, & Choragwicka, 2010; Peugh & Enders, 2004; Roth, 1994). Under missing data conditions, it is necessary for the data analyst to choose from among several available missing data techniques, including (a) listwise deletion, (b) pairwise deletion, (c) single imputation, (d) maximum likelihood (ML), and (e) multiple imputation (MI) approaches. That is, when facing missing data, abstinence is not an option-one of these missing data techniques must be used, and all are imperfect. Thus, the issue of selecting a missing data technique is a matter of choosing the lesser of evils.

Original languageEnglish (US)
Title of host publicationMore Statistical and Methodological Myths and Urban Legends
PublisherTaylor and Francis
Number of pages29
ISBN (Electronic)9781135039431
ISBN (Print)9780415838986
StatePublished - Jan 1 2014

ASJC Scopus subject areas

  • General Psychology


Dive into the research topics of 'Missing data bias: Exactly how bad is pairwise deletion?'. Together they form a unique fingerprint.

Cite this