Subgroup Analysis for Longitudinal Data via Semiparametric Additive Mixed Effects Model

This paper proposed a general framework based on semiparametric additive mixed effects model to identify subgroups on each covariate and estimate the corresponding regression functions simultaneously for longitudinal data,thus it could reveal which covariate contributes to the existence of subgroups among population.A backfitting combined with k-means algorithm was developed to detect subgroup structure on each covariate and estimate each semiparametric additive component across subgroups.A Bayesian information criterion is employed to estimate the actual number of groups.The efficacy and accuracy of the proposed procedure in identifying the subgroups and estimating the regression functions are illustrated through numerical studies.In addition,the authors demonstrate the usefulness of the proposed method with applications to PBC data and Industrial Portfolio's Return data and provide meaningful partitions of the populations.

Diagnosis of building energy consumption in the 2012 CBECS data using heterogeneous effect of energy variables:A recursive partitioning approach

Numerous previous literature has attempted to apply machine learning techniques to analyze relationships between energy variables in energy consumption.However,most machine learning methods are primarily used for prediction through complicated learning processes at the expense of interpretability.Those methods have difficulties in evaluating the effect of energy variables on energy consumption and especially capturing their heterogeneous relationship.Therefore,to identify the energy consumption of the heterogeneous relationships in actual buildings,this study applies the MOdel-Based recursive partitioning(MOB)algorithm to the 2012 CBECS survey data,which would offer representative information about actual commercial building characteristics and energy consumption.With resultant tree-structured subgroups,the MOB tree reveals the heterogeneous effect of energy variables and mutual influences on building energy consumptions.The results of this study would provide insights for architects and engineers to develop energy conservative design and retrofit in U.S.office buildings.

Personalised treatment assignment maximising expected benefit with smooth hinge loss

In personalised medicine,the goal is tomake a treatment recommendation for each patient with a given set of covariates tomaximise the treatment benefitmeasured by patient’s response to the treatment.In application,such a treatment assignment rule is constructed using a sample training data consisting of patients’responses and covariates.Instead of modelling responses using treatments and covariates,an alternative approach is maximising a response-weighted target function whose value directly reflects the effectiveness of treatment assignments.Since the target function involves a loss function,efforts have been made recently on the choice of the loss function to ensure a computationally feasible and theoretically sound solution.We propose to use a smooth hinge loss function so that the target function is convex and differentiable,which possesses good asymptotic properties and numerical advantages.To further simplify the computation and interpretability,we focus on the rules that are linear functions of covariates and discuss their asymptotic properties.We also examine the performances of our method with simulation studies and real data analysis.