Constrained maximum likelihood estimation of two-level covariance structure model via EM type algorithms

In this paper, the constrained maximum likelihood estimation of a two-level covariance structure model with unbalanced designs is considered. The two-level model is reformulated as a single-level model by treating the group level latent random vectors as hypothetical missing-data. Then, the popular EM algorithm is extended to obtain the constrained maximum likelihood estimates. For general nonlinear constraints, the multiplier method is used at theM-step to find the constrained minimum of the conditional expectation. An accelerated EM gradient procedure is derived to handle linear constraints. The empirical performance of the proposed EM type algorithms is illustrated by some artifical and real examples.This research was supported by a Hong Kong UCG Earmarked Grant, CUHK 4026/97H. We are greatly indebted to D.E. Morisky and J.A. Stein for the use of their AIDS data in our example. We also thank the Editor, two anonymous reviewers, W.Y. Poon and H.T. Zhu for constructive suggestions and comments in improving the paper. The assistance of Michael K.H. Leung and Esther L.S. Tam is gratefully acknowledged.     

Using finite mixture of GLMs to explore variability in children's flexibility in a task-switching paradigm

The present study illustrates the usefulness of finite mixture of generalized linear models (GLMs) to examine variability in cognitive strategies during childhood. More precisely, it addresses this variability in set-shifting situations where task-goal updating is endogenously driven. In a task-switching paradigm 5–6-year-olds had to switch between color- and shape-matching rules as a function of a predetermined, predictable task sequence. A finite mixture of GLMs was fitted to explore individual differences in performance. The statistical model revealed five response profiles, defined by accuracy and response times. These response profiles likely correspond to different cognitive strategies with varying efficiency and differential relations to working memory capacity (assessed by backward digit span). These results illustrate the heuristic value of statistical modeling to reveal the behavioral and cognitive variability in the temporal dynamics of children's cognitive functioning.     

The sandwich model: The ‘music and dance’ of therapeutic action

My premise is that a ‘layered’ approach is necessary to understand the process of exchanges that result in therapeutic change. I imagine these processes occurring in three layers – although the number of domains in which change is taking place is actually infinite – such as in a sandwich. The top layer, or top slice of bread of the sandwich, represents a broad view of the change process; it is non‐linear and includes the feature of uncertainty, a general principle of dynamic systems theory. The middle layer, or the meat of the sandwich, is explained by theories that are immediately and clinically useful to a therapist, such as psychoanalytic theories. These are primarily linear theories and use language and symbols to ‘tell a story of what happened’. The bottom layer, or bottom slice of bread of the sandwich, is the micro‐process; this layer includes the moment‐to‐moment patterns of coordinated rhythms that both communicate meaning and provide the essential scaffold for all higher‐level change processes. The micro‐process also requires a non‐linear theory to make sense of its variability and emergent properties. Taking a bite out of the sandwich will include a ‘polysemic bundle of communicative behaviors’ (Harrison and Tronick, 2011). I will illustrate the ‘sandwich model’ with the clinical case of the analytic treatment of a 5 year‐old boy.     

Levels of analysis and sources of data in sales research: a multilevel-multisource review

Sales researchers are increasingly adopting a multilevel-multisource (MLMS) approach to answer many key questions involving sales managers, salespeople and customers. MLMS research involves the acquisition and analysis of data collected from two or more sources pertaining to multiple hierarchical levels and presents a number of opportunities and challenges for sales researchers to consider. The authors highlight the benefits and the drawbacks of MLMS research and describe various approaches for implementing an MLMS collection and analysis. Additionally, a review of the MLMS studies conducted in the sales domain is provided which summarizes and delineates relationships tested in the extant literature. Based on this review, the authors advance a number of underdeveloped areas of research where MLMS approaches can be applied to further the understanding of the dynamic conditions that typify sales research.     

3～6岁儿童人格发展的群组序列追踪研究

目的:运用自编幼儿人格发展教师评定问卷对幼儿人格进行追踪测量,探讨其年龄及性别发展特点。方法:用整群抽样法选取3~3.5岁、3.5~4岁、4~4.5岁3个年龄群组幼儿为被试,采用群组序列的追踪设计,进行为期1年半的追踪测量,结合潜变量增长曲线模型和多层线性模型处理数据,探讨幼儿在3~6岁间人格的发展特点。结果:(1)幼儿的智能特征、认真自控、外倾性、亲社会性、情绪稳定性5个人格维度在3~4岁发展最快,4~5岁持续增长但发展速度放缓,到5~6岁时趋于平稳;(2)女孩的认真自控和亲社会性在3岁时显著高于男孩,但在3~6岁间的增长率不存在差异,即女孩的认真自控和亲社会性在幼儿阶段发展水平始终高于男孩。结论:从家庭进入幼儿园的环境变迁促使了幼儿人格的进一步发展,5岁左右幼儿人格开始初步形成;女孩的认真自控和亲社会性水平在幼儿阶段始终高于男孩。     

On the statistical and theoretical basis of signal detection theory and extensions: Unequal variance, random coefficient, and mixture models

Basic results for conditional means and variances, as well as distributional results, are used to clarify the similarities and differences between various extensions of signal detection theory (SDT). It is shown that a previously presented motivation for the unequal variance SDT model (varying strength) actually leads to a related, yet distinct, model. The distinction has implications for other extensions of SDT, such as models with criteria that vary over trials. It is shown that a mixture extension of SDT is also consistent with unequal variances, but provides a different interpretation of the results; mixture SDT also offers a way to unify results found across several types of studies.     

Linear instability signals the initiation of motion of a twin plane under load

This letter presents a study of the atomic mechanism of the initiation of motion of a static twin plane under applied mechanical load in a model shape-memory material. By tracking the deformation under load and using linear stability analysis, we find that the eigenvalues of the Hessian matrix provide an indicator of the initiation of motion of the twin plane. The initiation of motion is signaled by a linear instability and a drop in the lowest eigenvalue to zero as well as a sharp drop in higher eigenvalues. Additionally, by comparing with direct molecular dynamics, we see that the eigenmode associated with the zero eigenvalue is found to accurately predict the initial mode of motion. We also find that the initial motion occurs through the formation of a stacking fault just ahead of the existing twin plane and the broadening of the stacking fault drives further transformation.     

《心理统计学》教学内容的新探索

本文在对当前国内外主要心理统计学教材进行比较的基础上,指出与上个世纪八十年代的心理统计学教材内容相比较,在内容上的新探索主要体现在（1）由“假设检验”的内容中发展出“统计检验力”和“效果大小”的统计指标和估计方法;（2）引进一般线性模型来统合方差分析和回归分析这两种统计方法;（3）适度增加一些“多元统计分析”的内容等三个方面.本文对前两个方面的新内容作了简要评述,并对教材内容的编排方面提出了新的思路.     

文本结构和时间应激对网页阅读绩效的影响

探讨了文本结构和时间应激对网页阅读绩效的影响效应。采用 2× 3被试间设计 ,自变量为文本结构和阅读时间 ,因变量为找到相关信息的时间及途径的额外节点数。结果发现 ,主效应及交互作用的影响均十分显著。其中 ,当阅读时间为 10和 2 0min时 ,超文本阅读的绩效均显著低于线性文本 ,而在 30min时两者无显著差异 ;对超文本 ,三种阅读时间的效应存在显著差异 ,而对线性材料 ,阅读时间的效应无显著差异。结论 :时间应激对超文本阅读绩效的影响更为明显