GENetic EStimation and Inference in Structured samples (GENESIS): Statistical methods for analyzing genetic data from samples with population structure and/or relatedness
The fitted.values element of null models is different for non-mixed models and mixed-models. This had caused errors in SPA p-values for models without variance components or where variance components were estimated to be 0 (see #84).
Consider a bigger fix to make fitted.values consistent for all types of null model (e.g., it is always the same thing) but that is a much more involved change and has some issues with backwards compatibility (e.g., using a version of the null model run with an older version of GENESIS to do Score.SPA association tests in a new version).
The
fitted.valueselement of null models is different for non-mixed models and mixed-models. This had caused errors in SPA p-values for models without variance components or where variance components were estimated to be 0 (see #84).Consider a bigger fix to make
fitted.valuesconsistent for all types of null model (e.g., it is always the same thing) but that is a much more involved change and has some issues with backwards compatibility (e.g., using a version of the null model run with an older version of GENESIS to do Score.SPA association tests in a new version).Originally posted by @amstilp in https://github.com/UW-GAC/GENESIS/issues/84#issuecomment-1043468964