有监督Group MCP方法的稳健性研究

采用模拟研究的方法,分别在回归预测和分类判别两种环境中讨论有监督Group MCP方法在不同结构错误率下进行变量选择和结果预测的稳健性,并通过实例分析讨论本研究的实用价值。研究结果显示:忽略解释变量的内部结构进行变量选择会导致很多重要解释变量被疏漏,而有监督Group MCP方法考虑了解释变量的内部结构,在结构错误率低于5%时会以不低于98%的概率选出有效解释变量,并尽量降低冗余变量被选择的可能性。此研究成果为有监督Group MCP方法的合理使用奠定了基础。     

Constructing case vignettes for evaluating clinical judgment: An empirical model

Studies of clinical judgment are frequently based on analogue research, which uses experimenter-controlled stimuli to elicit judgments. The stimuli may be live models, audio- or video-taped models, or written case vignettes describing a clinical encounter. A major challenge of analogue research is to construct stimuli that maintain a balance between experimental rigor and clinical reality. An ideal set of case vignettes, for example, will contain summaries that resemble actual case histories and that are varied only on the specific clinical factors being studied. The model presented here demonstrates the empirical development of written case analogues in which several variables are studied simultaneously. The model can be adapted to any setting in which professionals are required to make judgments or decisions about individuals. The vignettes described here have been used in a variety of clinical settings to assess the reliability of clinicians'judgments and to aid in evaluation and program planning.     

Paths Following Algorithm for Penalized Logistic Regression Using SCAD and MCP

In this article, we develop a generalized penalized linear unbiased selection (GPLUS) algorithm. The GPLUS is designed to compute the paths of penalized logistic regression based on the smoothly clipped absolute deviation (SCAD) and the minimax concave penalties (MCP). The main idea of the GPLUS is to compute possibly multiple local minimizers at individual penalty levels by continuously tracing the minimizers at different penalty levels. We demonstrate the feasibility of the proposed algorithm in logistic and linear regression. The simulation results favor the SCAD and MCP’s selection accuracy encompassing a suitable range of penalty levels.     

Automatic structure discovery for varying-coefficient partially linear models

Varying-coefficient partially linear models provide a useful tools for modeling of covariate effects on the response variable in regression. One key question in varying-coefficient partially linear models is the choice of model structure, that is, how to decide which covariates have linear effect and which have non linear effect. In this article, we propose a profile method for identifying the covariates with linear effect or non linear effect. Our proposed method is a penalized regression approach based on group minimax concave penalty. Under suitable conditions, we show that the proposed method can correctly determine which covariates have a linear effect and which do not with high probability. The convergence rate of the linear estimator is established as well as the asymptotical normality. The performance of the proposed method is evaluated through a simulation study which supports our theoretical results.