We have implemented a strata analysis feature in PRSice-2 to aid the calculation of relative phenotypic risk of individuals between strata. Briefly, the N individuals of the target sample are first aggregated into M different strata based on their PRS. An N x (M − 1) design matrix is then generated using dummy coding, such that an individual is coded 1 in the column that corresponds to their PRS stratum and whereby a user-defined stratum is the reference group (or the median stratum by default). A linear regression (for quantitative traits) or logistic regression (for binary traits) will then be performed to estimate the phenotypic difference or relative risk, respectively, of each stratum vs the reference. The set of corresponding β-coefficients (linear) or the odds ratio (logistic) can then be visualized with the strata plot (Fig. 1). This allow users to assess whether individuals in the extreme stratum have a substantially higher phenotypic risk when compared to the reference stratum.
