The desire to maximize power also underlies the two reasons routinely cited to justify controlling for sex instead of directly assessing it. Not surprising, the main rationale is not enough statistical power to assess sex differences because the sample size has to be split in half. While it is often true that stratifying the data by sex will reduce statistical power, it is also true that power to detect an effect does not rely on sample size alone. Stratifying analyses by sex can result in increased power to detect significant genetic signals if the signals are small (or null) in one sex and larger in the other, or if large signals exist with opposite effects across sex [35]. In other words, if effect size is larger when split by sex than when sex is combined, statistical power may actually increase. Thus, the rationale of “not enough statistical power” becomes less justifiable.
