Issues in Statistical Inference--论文代写范文精选
2016-01-30 来源: 51due教员组 类别: Paper范文
Being critical of using significance tests in empirical research, the Board of Scientific Affairs (BSA) of the American Psychological Association (APA) convened a task force "to elucidate some of the controversial issues surrounding applications of statistics including significance testing and its alternatives; alternative underlying models and data transformation; and newer methods made possible by powerful computers" (BSA; quoted in the report by Wilkinson & Task Force, 1999, p. 594). Guidelines are stipulated in the report for revising the statistical sections of the APA Publication Manual. Some assertions in the report about research methodology are reasonable. An example is the statement, "There are many forms of empirical studies in psychology, including case reports, controlled experiments, quasi-experiments, statistical simulations, surveys, observational studies, and studies of studies (meta-analyses) ... each form of research has its own strengths, weaknesses, and standard of practice" (Wilkinson & Task Force, 1999, p. 594).
However, it does not follow that data collected with any two methods are equally unambiguous. At the same time, a method that yields less ambiguous data is methodological superior to one that yields more ambiguous data. That is, despite the assertions made in the report, a case can be made that "some of these [research methods] yield information that is more valuable or credible than others" (Wilkinson & Task Force, 1999, p. 594). It is unfortunate that the report reads more like an advocacy document than an objective assessment of the role of statistics in empirical research. Moreover, non-psychologist readers of the report can be excused for having a low opinion of psychologists' research practice and methodological sophistication.
Lest psychologists' methodological competence be misunderstood because of the report, this commentary addresses the following substantive issues: (a) the acceptability of the 'convenience' sample, (b) the inadequacy of the contrast group, (c) the unwarranted belief in the experimenter's expectancy effects, (d) some conceptual difficulties with effect size and statistical power, and (e) the putative dependence of statistical significance on sample size.
The 'Convenience' Sample, Representativeness and Independence of Observations
If we can neither implement randomization nor approach total control of variables that modify effects (outcomes), then we should use the term "control group" cautiously. In most of these cases, it would be better to forgo the term and use "contrast group" instead. In any case, we should describe exactly which confounding variables have been explicitly controlled and speculate about which unmeasured ones could lead to incorrect inferences.
In the absence of randomization, we should do our best to investigate sensitivity to various untestable assumptions. (Wilkinson & Task Force, 1999, p. 595, emphasis in italics added) A non-randomly selected sample is characterized as a "convenience sample" (Wilkinson & Task Force, 1999, p.595). It is a label apparently applicable to most samples used in psychological research because most experimental subjects are college student-volunteers. However, a case can be made that using such non-random samples does not necessarily detract from the findings generality. Nor does such a practice violate the requirement that data fromdifferent subjects be statistically independent. More importantly, using non-random samples is not antithetical to experimental controls.
Suppose that, on the basis of the data collected from student-subjects, Experimenter E draws a conclusion about memory. The non-random nature of the sample would not affect the objectivity of the finding when the validity of the experiment is assessed with reference to unambiguous, theoretically informed criteria. At worst, one may question the generality of the experimental conclusion. Perhaps, this is the real point of the "Sample" section (Wilkinson & Task Force, 1999, p. 595), as witnessed by its reservations about the representativeness of the convenience sample. Although non-random selection of research participants jeopardizes the generality of survey studies, random subject-selection may not be necessary for generality in cognitive psychology. For instance, a non-random sample in an opinion survey about an election may be selected by stationing the enumerators at the entrance of a shopping mall. The representativeness of the opinion of such a sample (of the entire electorate's opinion) is suspect because patrons of the particular shopping mall may over-represent one social group, but under-represent another social strata.
This is crucial because political opinion and socio-economic status are not independent. In contrast, consider a student-subject sample of a study of the capacity of the short-term store. As there is no reason to doubt the similarity between college students' short-term store capacity and that of the adult population at large, it is reasonable to assume that the student-subject sample is representative of all adults in the said capacity despite that no random selection is carried out. That is, random selection is not always required for establishing the generality of the result when there is neither a theoretical nor an empirical reason to question the representativeness of the sample in the context of the experiment.
The psychologist's practice of using student-subjects is further justified by the fact that psychologists employ student-subjects in a theoretically informed way. For example, in testing a theory about verbal coding, the experimenter may use only female students. The experimenter may use only right-handed students when the research concern is a theory about laterality or hemispheric specialization. Students may be screened with the appropriate psychometric tests before being included in a study about attitude. In short, depending on the theoretical requirement, psychologists adopt special subject-selection criteria even when they use studentsubjects. Moreover, psychologists do select subjects from outside the student-subject pools when required (e.g., they use hyperactive boys to study theories of hyperactivity). The mode of subject-selection is always explicitly described in such an event. That is, psychologists' convenience samples do not detract from the data's generality. Furthermore, psychologists describe only those procedural features that deviate from the usual, well-understood and warranted practice.
A crucial assumption underlying statistical procedures (be it significance test, confidence-interval estimate or regression analysis) is that observations are independent of one another. It can be illustrated that cognitive psychologists' use of non-randomly selected student-subjects does not violate this independence assumption. Consider the case in which, having discussed among themselves, twenty students decide to participate in the same memory experiment. This is non-random subject-selection par excellence. Suppose further that subjects, whose task is to recall multiple 10-word lists in the order they are presented, are tested individually.
The words and their order of appearance are randomized from trial to trial. Under such circumstances, not only would an individual subject's performance be independent of that of other subjects, the subject's performance is also independent of his or her own performance from list to list. In other words, to ensure statistical independence of observations, what needs to be randomized is the stimulus material or its mode of presentation, not individual subjects. Such a randomized procedure ensures that non-randomly selected subjects may still produce statistically independent data.
any causal relationship is based on the implicative relationships among the explanatory theory, the research hypothesis, the experimental hypothesis, the statistical hypothesis, and the data (see, e.g., the three embedding conditional syllogisms discussed in Chow, 1996, 1998). The causal conclusion owes its ambiguity to deductive logic as a result of the facts that (a) hypothetical properties are attributed to the unobservable theoretical entities postulated (Feigl, 1970; MacCorquodale & Meehl, 1948), (b) it is always possible to offer multiple explanations for the same phenomenon (Popper, 1968a, 1968b), and (c) affirming the consequent of a conditional proposition does not affirm its antecedent (Cohen & Nagel, 1934; Meehl, 1967, 1978). In other words, the report's treatment of random subject-assignment is not helpful when it incorrectly assigns to the research design the task of making causal inference possible. Nor is the ambiguity of drawing causal conclusions a difficulty in inductive logic, as said in the report that "the causal inference problem ... one of missing data" (Wilkinson & Task Force, 1999, p.600).(paper代写)
更多paper代写范文欢迎访问我们主页 www.51due.com 当然有paper代写需求可以和我们24小时在线客服 QQ:800020041 联系交流。-X(paper代写)