粗糙集和神经网络在心理测量中的应用

探讨当因素分析和多元回归方法的使用条件未得到满足时,是否可采用粗糙集方法进行观察变量的精简,以及是否可采用神经网络方法进行预测效度检验。理论分析了粗糙集和神经网络在心理测量中应用的可能性,并运用粗糙集对于人事干部胜任力评估数据进行分析,比较了7种离散化方法和2种约简算法构成的14种组合,发现当采用Manual方法进行离散化、遗传算法进行约简时,能够很好地对观测变量进行精简;运用概率神经网络能够比等级回归方法更好地进行预测效度检验。研究结果表明对于处理心理测量中的非等距变量,粗糙集和神经网络是非常有用的方法     

Contributions of Psychophysiology to Media Research: Review and Recommendations

Despite the increasing use of psychophysiological measures in various research areas, there is a relative paucity of studies on communication, media, and media interfaces that have taken advantage of this approach. This article provides an overview of the use of psychophysiological measures of attention and emotion in media research with the focus on 3 most commonly used measures: heart rate, facial electromyography, and electrodermal activity. Selected media studies that have used psychophysiological methods to test theory-based predictions regarding the role of attentional and emotional factors in message processing are critically reviewed. The article also highlights some methodological and other issues critical for the successful application of psychophysiological methods to problems in media research. In particular, respiratory sinus arrhythmia (RSA), a selective index of parasympathetic nervous system activity, is introduced as a measure that holds particular promise for media research, given that RSA is highly sensitive to changes in attention.     

Retrieving relevant factors with exploratory SEM and principal-covariate regression: A comparison

Behavioral researchers often linearly regress a criterion on multiple predictors, aiming to gain insight into the relations between the criterion and predictors. Obtaining this insight from the ordinary least squares (OLS) regression solution may be troublesome, because OLS regression weights show only the effect of a predictor on top of the effects of other predictors. Moreover, when the number of predictors grows larger, it becomes likely that the predictors will be highly collinear, which makes the regression weights’ estimates unstable (i.e., the “bouncing beta” problem). Among other procedures, dimension-reduction-based methods have been proposed for dealing with these problems. These methods yield insight into the data by reducing the predictors to a smaller number of summarizing variables and regressing the criterion on these summarizing variables. Two promising methods are principal-covariate regression (PCovR) and exploratory structural equation modeling (ESEM). Both simultaneously optimize reduction and prediction, but they are based on different frameworks. The resulting solutions have not yet been compared; it is thus unclear what the strengths and weaknesses are of both methods. In this article, we focus on the extents to which PCovR and ESEM are able to extract the factors that truly underlie the predictor scores and can predict a single criterion. The results of two simulation studies showed that for a typical behavioral dataset, ESEM (using the BIC for model selection) in this regard is successful more often than PCovR. Yet, in 93% of the datasets PCovR performed equally well, and in the case of 48 predictors, 100 observations, and large differences in the strengths of the factors, PCovR even outperformed ESEM.     

Latent variables should remain as such: Evidence from a Monte Carlo study

Use of subject scores as manifest variables to assess the relationship between latent variables produces attenuated estimates. This has been demonstrated for raw scores from classical test theory (CTT) and factor scores derived from factor analysis. Conclusions on scores have not been sufficiently extended to item response theory (IRT) theta estimates, which are still recommended for estimation of relationships between latent variables. This is because IRT estimates appear to have preferable properties compared to CTT, while structural equation modeling (SEM) is often advised as an alternative to scores for estimation of the relationship between latent variables. The present research evaluates the consequences of using subject scores as manifest variables in regression models to test the relationship between latent variables. Raw scores and three methods for obtaining theta estimates were used and compared to latent variable SEM modeling. A Monte Carlo study was designed by manipulating sample size, number of items, type of test, and magnitude of the correlation between latent variables. Results show that, despite the advantage of IRT models in other areas, estimates of the relationship between latent variables are always more accurate when SEM models are used. Recommendations are offered for applied researchers.     

Are psychophysiological/psychophysical scores good candidates for diagnosing temperament/personality traits and for a demonstration of the construct validity of psychometrically measured traits?

The aim of this paper is to examine the use of psychophysiological and/or psychophysical measures employed as (a) tools aimed at diagnosing personality/temperament traits and (b) criteria for estimating the construct validity of personality/temperament inventories. The exploration is based on three diferent sources: (1) data collected mainly within the neo-Pavlovian approach on CNS properties; (2) selected examples of studies in the domain of sensation-seeking, neuroticism, and extraversion; and (3) theoretical considerations regarding the specificity of phenomena being measured by means of psychometric and psychophysiological/psychophysical scores. It is concluded that psychophysiological and psychophysical measures may be used for the assessment of personality/temperament traits or for estimating the construct validity of psychometric tools applied in personality (temperament) research under very strict requirements and limited circumstances only. In most of the studies conducted up to now, these circumstances and requirements have not been fulfilled.     

Examining classification criteria: a comparison of three cut score methods

This study compared 3 different methods of creating cut scores for a screening instrument, T scores, receiver operating characteristic curve (ROC) analysis, and the Rasch rating scale method (RSM), for use with the Behavioral and Emotional Screening System (BESS) Teacher Rating Scale for Children and Adolescents (Kamphaus & Reynolds, 2007). Using the BESS norm data set, we compared the methods across 7 classification indices. Additional information about accuracy was used with a subset of children who had been given a prior diagnosis for selected disorders. The results showed that the methods were generally in concordance, with similarities identified across methods. RSM and ROC analysis methods performed similarly, with both methods identifying the same optimal cut-point. The method based on T scores appeared to be more conservative, identifying a lower cut score as optimal.     

Psychometric Modeling of response speed and accuracy with mixed and conditional regression

Human performance in cognitive testing and experimental psychology is expressed in terms of response speed and accuracy. Data analysis is often limited to either speed or accuracy, and/or to crude summary measures like mean response time (RT) or the percentage correct responses. This paper proposes the use of mixed regression for the psychometric modeling of response speed and accuracy in testing and experiments. Mixed logistic regression of response accuracy extends logistic item response theory modeling to multidimensional models with covariates and interactions. Mixed linear regression of response time extends mixed ANOVA to unbalanced designs with covariates and heterogeneity of variance. Related to mixed regression is conditional regression, which requires no normality assumption, but is limited to unidimensional models. Mixed and conditional methods are both applied to an experimental study of mental rotation. Univariate and bivariate analyzes show how within-subject correlation between response and RT can be distinguished from between-subject correlation, and how latent traits can be detected, given careful item design or content analysis. It is concluded that both response and RT must be recorded in cognitive testing, and that mixed regression is a versatile method for analyzing test data.I am grateful to Rogier Donders for putting his data at my disposal.     

CHull: A generic convex-hull-based model selection method

When analyzing data, researchers are often confronted with a model selection problem (e.g., determining the number of components/factors in principal components analysis [PCA]/factor analysis or identifying the most important predictors in a regression analysis). To tackle such a problem, researchers may apply some objective procedure, like parallel analysis in PCA/factor analysis or stepwise selection methods in regression analysis. A drawback of these procedures is that they can only be applied to the model selection problem at hand. An interesting alternative is the CHull model selection procedure, which was originally developed for multiway analysis (e.g., multimode partitioning). However, the key idea behind the CHull procedure—identifying a model that optimally balances model goodness of fit/misfit and model complexity—is quite generic. Therefore, the procedure may also be used when applying many other analysis techniques. The aim of this article is twofold. First, we demonstrate the wide applicability of the CHull method by showing how it can be used to solve various model selection problems in the context of PCA, reduced K-means, best-subset regression, and partial least squares regression. Moreover, a comparison of CHull with standard model selection methods for these problems is performed. Second, we present the CHULL software, which may be downloaded from http://ppw.kuleuven.be/okp/software/CHULL/, to assist the user in applying the CHull procedure.     

Functional measurement approach to self-estimation in multiattribute evaluation

Multiattribute analysis depends on measurement of values and weights. Unless these measures reflect the decision maker's true values and weights, the multiattribute formula may put a less-preferred alternative in first place. To avoid such disordinality requires stringent measurement conditions: First, the values and weights must be on linear (equal interval) or ratio (known zero) scales. Second, these scales must satisfy a condition of common unit across disparate attribute dimensions. Most methods of range adjustment beg both of these measurement questions. Functional measurement theory can solve both problems and so can be useful in multiattribute analysis. Past work has established the operation of a general cognitive algebra as an empirical reality. The averaging model, in particular, makes possible the definition and estimation of weights and values as distinct psychological parameters. It can also solve the problem of common unit. Cognitive algebra thus provides a grounded theoretical foundation on which to develop self-estimation methodology, in which decision makers provide direct estimates of their values and weights. The logic is straightforward. Functional measurement can analyze global judgments to obtain validated psychological scales. These scales may then be used as validational criteria for the self-estimates. Procedures to eliminate biases in the self-estimates can thus be tested and refined in well-learned multiattribute tasks, such as judgments of meals, in which global judgments are trustworthy. Once developed, such self-estimation procedures may be used with some confidence for general multiattribute analysis. A number of studies from 20-odd years of work on the theory of information integration are summarized to show good, although not unmixed promise for self-estimation.     

METHODS OF IMPROVING EMPLOYMENT TEST CRITICAL SCORES DERIVED BY JUDGING TEST CONTENT: A REVIEW AND CRITIQUE

Setting critical scores on content valid tests may involve making judgments about test content. A conclusion which has been reached in the testing literature is that the Angoff method is preferred as this type of judgmental method of setting standards or critical scores. The current paper reflects this conclusion, but also discusses several limitations of the Angoff method, and outlines why less than optimal data may be obtained using this preferred method or similar methods. Several techniques are reviewed and critiqued that may be used in conjunction with the Angoff method or similar methods to attempt improvement of the psychometric quality of the critical score data. Components from two psychometric frameworks (generalizability theory and Cronbach's, 1955, accuracy scores) are integrated and applied in assessing the impact that each of the techniques may have on Angoff critical score data. The techniques reviewed can be roughly divided into: (a) procedural methods (involved in the judgment process), and (b) psychometric methods (involved in data analysis following judgment). The present review is intended as a resource to personnel testing specialists who may be interested in methods of bolstering the process of judgmentally deriving critical scores.     

Psychophysiological arousal related to Type A components in adolescent boys

The association between psychophysiological responses (heart rate, skin conductance and blood volume) and Type A behavior was studied in adolescent boys (n = 48) in computer-controlled experiments. Although psychophysiological arousal was related to the type of stress-evoking element, task-specificity did not result in significant psychophysiological differences between Type As and Nontype As. The indication is that physiological arousal may be a constitutional characteristic of Type A behavior. The multidimensionality of type A behavior must be considered in any investigation examining the psychophysiological Type A-Nontype A differences. Different Type A dimensions, together with previously found psychological differences, were related to specific psychophysiological reactions.     

GT与IRT的比较: 北京奥运会男子10米跳台跳水分析

概化理论(GT)和项目反应理论(IRT)从两个不同的方向发展了经典测量理论, GT和IRT中的多面Rasch测量模型(MFRM)在主观评分中都可以用来估计评分中各变异来源对变异的贡献, 对测评的信度进行估计, 提出测评改进意见。12名运动员参加了2008北京奥运会男子10米跳台跳水决赛, 比赛共6个回合, 7名裁判独立对他们在各个回合的表现进行打分。GT和MFRM比较一致地认为运动员自身、回合、运动员与回合的交互效应是运动员得分的重要变异来源, 而裁判员对运动员得分差异的贡献不显著。MFRM同时还估计出难度系数是影响男子跳台跳水成绩的重要变异来源, 在评分等级6.5附近存在步校准错乱, 得出的运动员成绩排序与2008奥运实际排序有所不同。在GT中难度系数作为隐藏侧面, 其效应未能分离出来。GT和MFRM从两个不同的方面给测量提供改进意见: GT发现可以通过增加回合数来提高g系数, 而增加裁判数对其影响不大。MFRM给出各侧面的要素(如某裁判、运动员等)的估计值及其标准误, 它给出的诊断性拟合统计也有助于甄别异常得分或评分模式。     

The efficacy of cue exposure with response prevention in extinguishing drug and alcohol cue reactivity

It has been suggested that drug cue‐elicited urges and psychophysiological reactions are the results of Pavlovian conditioning processes and that it should be possible to extinguish these responses with cue exposure with response prevention. It has already been shown that subjective cue‐elicited urges can be extinguished, but it is unclear whether this is also true for cue‐elicited psychophysiological arousal. This was tested in the present study in a heterogeneous sample of drug and alcohol dependent patients. It was found that cue‐elicited urges can indeed be extinguished. However, such a clear pattern of extinguished cue reactivity was not found for the psychophysiological measures. Furthermore, the extinction of drug urges was not specific for cue exposure treatment. It is concluded that cue‐elicited psychophysiological arousal does not underlie subjective cue reactivity and may not reflect Pavlovian conditioned drug responding. Copyright © 2006 John Wiley & Sons, Ltd.     

基于遗传算法的模糊综合评价在心理测量中的应用

提出了运用模糊数学对利克特量表数据进行分析的方法, 探讨了人们在进行模糊综合评价时, 所采用的算子和对各个自变量的权重分配, 并且运用遗传算法(GA)来确定相关的权重。以大学生对康师傅红烧牛肉面的评价数据为例, 运用基于遗传算法的模糊综合评价方法, 发现男生采用了“最大最小”合成算子, 女生采用了“有界和、取小”合成算子。研究结果表明, 基于遗传算法的模糊综合评价方法可以对利克特量表的心理测量数据进行有效的分析。     

Predicting Latent Class Scores for Subsequent Analysis

Latent class regression models relate covariates and latent constructs such as psychiatric disorders. Though full maximum likelihood estimation is available, estimation is often in three steps: (i) a latent class model is fitted without covariates; (ii) latent class scores are predicted; and (iii) the scores are regressed on covariates. We propose a new method for predicting class scores that, in contrast to posterior probability-based methods, yields consistent estimators of the parameters in the third step. Additionally, in simulation studies the new methodology exhibited only a minor loss of efficiency. Finally, the new and the posterior probability-based methods are compared in an analysis of mobility/exercise.     

Quantile Mediation Models: A Comparison of Methods for Assessing Mediation Across the Outcome Distribution

Recent introduction of quantile regression methods to analysis of epidemiologic data suggests that traditional mean regression approaches may not suffice for some health outcomes such as Body Mass Index (BMI). In the same vein, the traditional mean-based approach to mediation modeling may not be sufficient to capture the potentially different mediating effects of behavioral interventions across the outcome distribution. By combining methods for estimating conditional quantiles with traditional mediation modeling techniques, mediation effects can be estimated for any quantile of the outcome distribution (so-called quantile mediation effects). Estimation and inference techniques for quantile mediation effects are compared through simulation studies, and recommendations are given. The quantile mediation methods are further compared with the traditional mean-based regression approaches to mediation analysis through analysis of data from Healthy Places, a trial that is examining the effects of the community–built environment on resident obesity risk. We found the magnitudes of indirect (mediating) effects of walkability on BMI and waist circumference were substantially larger for the upper quantiles compared with the median or mean. Results suggest that restricting the examination of mediation to the mean of the outcome distribution provides an incomplete picture of proposed mediating mechanisms and in some cases may miss important mediational relationships to outcomes.     

The Tower of London Test: different scoring criteria for diagnosing Alzheimer's disease and mild cognitive impairment

The Tower of London (TOL) is used for evaluating planning skills, which is a component of the executive functions. Different versions and scoring criteria were developed for this task, and some of them present with different psychometrical properties. This study aimed to evaluate two specific scoring methods of the TOL in diagnosing Mild Cognitive Impairment and probable Alzheimer's disease. The TOL total scores from 60 patients of each diagnosis were compared with the performance of 60 healthy-aged controls using receiver operating characteristics analysis and multinomial logistic regression. Krikorian method better diagnosed Alzheimer's disease, while Portellas's was better at discriminating healthy controls from Mild Cognitive Impairment, but were not efficient at comparing this last group with Alzheimer's patients. Regression analysis indicates that in addition to screening tests, TOL improves the classification of the three groups. The results suggest the two scoring methods used for this task may be useful for different diagnostic purposes.     

Bayesian estimation of a multilevel IRT model using gibbs sampling

In this article, a two-level regression model is imposed on the ability parameters in an item response theory (IRT) model. The advantage of using latent rather than observed scores as dependent variables of a multilevel model is that it offers the possibility of separating the influence of item difficulty and ability level and modeling response variation and measurement error. Another advantage is that, contrary to observed scores, latent scores are test-independent, which offers the possibility of using results from different tests in one analysis where the parameters of the IRT model and the multilevel model can be concurrently estimated. The two-parameter normal ogive model is used for the IRT measurement model. It will be shown that the parameters of the two-parameter normal ogive model and the multilevel model can be estimated in a Bayesian framework using Gibbs sampling. Examples using simulated and real data are given.     

The prediction of intelligence in preschool children using alternative models to regression

Statistical prediction of an outcome variable using multiple independent variables is a common practice in the social and behavioral sciences. For example, neuropsychologists are sometimes called upon to provide predictions of preinjury cognitive functioning for individuals who have suffered a traumatic brain injury. Typically, these predictions are made using standard multiple linear regression models with several demographic variables (e.g., gender, ethnicity, education level) as predictors. Prior research has shown conflicting evidence regarding the ability of such models to provide accurate predictions of outcome variables such as full-scale intelligence (FSIQ) test scores. The present study had two goals: (1) to demonstrate the utility of a set of alternative prediction methods that have been applied extensively in the natural sciences and business but have not been frequently explored in the social sciences and (2) to develop models that can be used to predict premorbid cognitive functioning in preschool children. Predictions of Stanford–Binet 5 FSIQ scores for preschool-aged children is used to compare the performance of a multiple regression model with several of these alternative methods. Results demonstrate that classification and regression treesprovided more accurate predictions of FSIQ scores than does the more traditional regression approach. Implications of these results are discussed.     

Heuristic cognitive diagnosis when the Q‐matrix is unknown

Cognitive diagnosis models of educational test performance rely on a binary Q‐matrix that specifies the associations between individual test items and the cognitive attributes (skills) required to answer those items correctly. Current methods for fitting cognitive diagnosis models to educational test data and assigning examinees to proficiency classes are based on parametric estimation methods such as expectation maximization (EM) and Markov chain Monte Carlo (MCMC) that frequently encounter difficulties in practical applications. In response to these difficulties, non‐parametric classification techniques (cluster analysis) have been proposed as heuristic alternatives to parametric procedures. These non‐parametric classification techniques first aggregate each examinee's test item scores into a profile of attribute sum scores, which then serve as the basis for clustering examinees into proficiency classes. Like the parametric procedures, the non‐parametric classification techniques require that the Q‐matrix underlying a given test be known. Unfortunately, in practice, the Q‐matrix for most tests is not known and must be estimated to specify the associations between items and attributes, risking a misspecified Q‐matrix that may then result in the incorrect classification of examinees. This paper demonstrates that clustering examinees into proficiency classes based on their item scores rather than on their attribute sum‐score profiles does not require knowledge of the Q‐matrix, and results in a more accurate classification of examinees.