Nonparametric variable screening for multivariate additive models

In this paper we develop a novel procedure of variable screening for multivariate additive random-effects model, based on B-spline function approximations. With these approximations, the so-called signal-to-noise ratio (SNR) can be defined to inform importance each covariate in model. Then, SNR-based forward filtering is conducted covariates by using iterative projections multiple response data into space covariates. The proposed easy use and allows user pool non-linear information across heterogeneous subjects through variables. We establish an asymptotic theory selection consistency under some regularity conditions. By simulations, show that has superior performance over existing methods terms sensitivity specificity. also apply anti-cancer drug data, revealing set biomarkers potentially influence concentrations drugs cancer cell lines.