Shape detection using semi-parametric shape-restricted mixed effects regression spline with applications (Q2047377)

Classical linear mixed models assume a linear relationship between responses and the matrix of fixed effects covariates, which may be restrictive in some statistical applications. Semi-parametric shape-restricted mixed effects regression spline models are proposed in this paper to capture potential monotone or convex functional relationship between the responses and some covariates. A two-step approach is proposed to estimate the parameters in the proposed models. Procedures are also proposed for the selection of the most reasonable shape of the functional relationship. Simulation studies are conducted to evaluate the family-wise type I error and power of the proposed procedures. The proposed procedures are also illustrated through the application to a prenatal screening program dataset for the inferences on the relationships between placental-fetal hormones and birth weight while taking into account of geographic locations of maternal residence as a random effect.