The seven deadly sins of statistical analysis.

Abstract

In a pedantic but playful way, we discuss some common errors in the use of 'statistical analysis' that are regularly observed in our professional plastic surgical literature. The seven errors we discuss are (1) the use of parametric analysis of ordinal data; (2) the inappropriate use of parametric analysis in general; (3) the failure to consider the possibility of committing type II statistical error; (4) the use of unmodified t-tests for multiple comparisons; (5) the failure to employ analysis of covariance, multivariate regression, nonlinear regression, and logistical regression when indicated; (6) the habit of reporting standard error instead of standard deviation; and (7) the underuse or overuse of statistical consultation. Confidence and common sense are advocated as a means to balance statistical significance with clinical importance.