Inconsistent use of multiple comparison corrections in studies of population genetic structure: Are some type I errors more tolerable than others?

Abstract

Studies of genetic population structure often involve numerous tests of Hardy–Weinberg equilibrium (HWE), linkage disequilibrium (LD) and genetic differentiation. Tests of HWE or LD are important precursors to population structure assessments. When conducting multiple related statistical tests, type I error increases, e.g., familywise error rate (FWER) inflation. FWER inflation can alter the results of statistical tests and thus the conclusions. We surveyed literature from 2011 to 2013 to determine if studies of population structure assess LD and HWE and if FWER corrections were applied consistently across different types of tests. We found significantly inconsistent FWER corrections, with a bias towards less restrictive correction on genetic differentiation and more restrictive corrections with LD and HWE. While varied adjustments of FWER for different types of analyses might be justified, papers with inconsistent usage across tests of HWE, LD and genetic differentiation did not present rationale for their FWER corrections. We also found a lack of documentation of HWE, LD and FWER corrections in studies. We encourage the authors to report statistical tests and related FWER corrections, use FWER corrections consistently or justify their different methods in the same study.