We demonstrate a method for localizing where two spline terms, or smooths, differ using a true discovery proportion (TDP) based interpretation. The procedure yields a statement on the proportion of some region where true differences exist between two smooths, which results from use of hypothesis tests on collections of basis coefficients parameterizing the smooths. The methodology avoids otherwise ad hoc means of making such statements like subsetting the data and then performing hypothesis tests on the truncated spline terms. TDP estimates are 1-alpha confidence bounded simultaneously. This means that the TDP estimate for a region is a lower bound on the proportion of actual difference, or true discoveries, in that region with high confidence regardless of the number of regions at which TDP is estimated. Our procedure is based on closed-testing using Simes local test. This local test requires that the `multivariate chi-sq test statistics of generalized Wishart type' underlying the method are positive regression dependent on subsets (PRDS), which we show. The method is well-powered because of a result on the off-diagonal decay structure of the covariance matrix of penalized B-splines of degree two or fewer. We demonstrate achievement of estimated TDP in simulation and analyze a study of walking gait of cerebral palsy patients.
翻译:暂无翻译