Domains of convergence for the EM algorithm: a cautionary tale in a location estimation problem

The EM algorithm is a popular method for maximizing a likelihood in the presence of incomplete data. When the likelihood has multiple local maxima, the parameter space can be partitioned into domains of convergence, one for each local maximum. In this paper we investigate these domains for the location family generated by the t-distribution. We show that, perhaps somewhat surprisingly, these domains need not be connected sets. As an extreme case we give an example of a domain which consists of an infinite union of disjoint open intervals. Thus the convergence behaviour of the EM algorithm can be quite sensitive to the starting point.     

A comparative approach to assessing the quality of life of intercollegiate athletes

Although concerns have frequently been raised recently regarding the quality of life of intercollegiate athletes, information is seldom available on which college and university administrators can base policy decisions. Particularly lacking are studies that provide administrators with assessments of their own athletics programs relative to athletics programs at other institutions. We describe here a method we used in conducting a comparative evaluation of the status of student-athletes at a large public university, involving comparisons of student-athletes at this university to student-athletes at a subset of the institutions that participated in the 1987–88 National Study of Intercollegiate Athletes. By utilizing existing data sets to construct comparison samples, relative assessments of the status of student-athletes at a given institution can be performed without the excessive costs of gathering data on student-athletes at comparison institutions. Techniques employed in data collection and analysis are discussed as well as the format of the report in which the findings of the study were presented.     

Progress and cumulation in the human sciences after the fall

What do we mean by progress and cumulation in the social and human sciences? Recent thinking in the philosophy and history of science has led to an abandonment of some versions of logical positivism and of verificationism that had a strong deductive and theory testing orientation. What is to replace them is less clear. This paper argues that progress and cumulation can be seen as a process of evaluation and retention within an epistemic community. Scholarly disciplines differ in their social structure and in their epistemic and normative commitments. Since sociology is a fragmented discipline, progress and cumulation differ within its multiple subdisciplines, which to varying extents represent epistemic communities. Brief sketches of progress (advance) and cumulation in several subdisciplines are offered.     

A theory of prudential values and a rule utilitarian theory of morality

Ethics can be divided into a theory of prudential values and a theory of morality in a narrower sense. My paper proposes a utilitarian — a rule-utilitarian — theory of morality. But it deviates from most of the utilitarian tradition by rejecting the hedonistic and subjectivistic accounts of prudential values favored by many utilitarian writers. While economists tend to define people's utility levels in terms of their actual preferences, ethics must define them in terms of their informed preferences. To prefer A over B does not mean to have a stronger desire for A than for B. Rather, it means to regard one's access to A as being more important than one's access to B. Even though different people often have quite different preferences, their basic desires seem to be much the same. We must choose our moral rules, and our society's moral code as a whole, by their social utility. An important factor in determining their social utility are their expectation effects. Unlike the rule — utilitarian more code, the act — utilitarian moral code would be unable to give proper weight to these expectation effects. It would also unduly restrict our individual freedom. Finally, I shall argue against Kant that morality is primarily a servant of many other human values rather than itself the highest value of human life.     

应用神经元控制器的实时控制装置和实验

针对平衡倒立摆这一非线性控制问题，设计了神经元控制器和控制装置，编制了实时控制程序．控制装置运行良好，成功地平衡了倒立摆．实时控制实验显示，当倒立摆的参数变化时，神经元控制器比一般的控制器具有强得多的自适应能力．     

几组复合麻醉方法对犬麻醉效果比较

复方846、静松灵/氯胺酮、氯丙嗪/氯胺酮、氯丙嗪/乙醚、氯丙嗪/静松灵,均可使犬获得一定时间的麻醉效果。通过麻醉前、后体温(T℃)、呼吸(R),脉搏(P),镇痛反应,麻醉维持时间的检测;通过麻醉后手术安静顺利情况;从伤口愈合情况,以及术前、术后3天、7天白细胞、淋巴细胞、T淋巴细胞的变化表明:几组药物作用于犬后,体温间差异不显著(P>0.05),呼吸、脉搏部分呈显著性变化(P>0.05),各组术前、术后3天、7天的白细胞、淋巴细胞、T淋巴细胞变化组间差异不显著(P>0.05)。镇痛作用以复方846为好,其次为静松灵/氯胺酮、氯丙嗪/静松灵、氯丙嗪/氯胺酮、氯丙嗪/乙醚;麻醉维持时间复方846最长(133.70±9.66),其次为静松灵/氯胺酮(89±9.9);从手术安静顺利情况看,复方846、静松灵/氯胺酮优于其它组;从伤口愈合情况看,静松灵/氯胺酮优于其它组。综合比较:复方846、静松灵/氯胺酮用于大的临床麻醉,效果确实。     

The psychodynamically oriented clinical social worker as sports consultant: A preliminary report

This paper explores an unusual and innovative application of psychodynamically oriented clinical social work method to the sports area. The coach of a national Olympics team requested a clinical consultation because he suspected that there might be psychosocial reasons why a certain number of world class atheletes he was coaching were not performing up to their full potential.     

Correlates of female crime

This article discusses the complexity of developing correlates of female crime in the comparative context of Western and Indian literature. It selectively reviews the findings of American and Indian studies of female crime and concludes with the author’s own findings from a study of 120 women prisoners conducted in a federal women’s prison in India. The author’s main contention is that whereas broad contours of offender profiles can be drawn, it is difficult to draw statistically significant relationships and infer from them predictable variables of criminality.