Characteristics of simulation models.

<p>*. The simulation model consists of 5 covariates, each with OR of 1.5. In analysis, we assume that only the last two covariates are considered.</p><p>$. In analysis, a more stringent threshold (10⁻⁴) is used for significance, compared to other simulation models (.05).</p><p>n: number of cases, assuming case:control ratio of 1∶1; RAF: risk allele frequency; OR_g: odds ratio for risk allele; OR_z: odds ratios for covariates; corr: correlation coefficient between causal variant and the last covariate; α_max: variation in sample proportions (see “<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0114523#s2" target="_blank">Methods</a>”). Model 1–4 were simulated under the null hypothesis of no association; and model 5–12 were under the alternative hypothesis. Model 1 and 5 were treated as baseline models, and changes of parameters in other models were highlighted.</p><p>Characteristics of simulation models.