Statistical practices: the seven deadly sins

Abstract

This paper discusses selected problems in applied statistical analysis: (a) over-reliance on null hypothesis statistical testing, (b) failing to perform a power analysis prior to conducting the study, (c) using asymptotic statistical approximations with small samples, (d) ignoring missing data, (e) failing to deal with the multiplicity problem when performing multiple statistical comparisons, (f) using stepwise procedures to select variables in regression analysis, and (g) failing to perform or report model diagnostics. Suggestions and guidelines to address these issues in manuscripts concerning child neuropsychological research are offered.