文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Bayesian inference with probability matrix decomposition models
侵权投诉

Bayesian inference with probability matrix decomposition models

check, psychometrics Probability Mcatrix Decompositioni models mtay bve uised to model observed binary associations between two sets of elements. More specifically, to explain observed associations betweeni two elements, it is assumed that B laitent Bernou

Bayesian Inference U'ith PMI) Models

PMD models assume a two-fold process to explain associations between objects and attributes: First, when judging the associationi between elements (objects and attributes), it is assumed that raters may (or rnay not) perceive each of a set of latentftatures in each element. Second, it is assumed that judgments regarding t

he association of two elements are the result of combining the feature patterns of these elements according to a specific rule. In general, all kinds of rules mnay be considered but in the literature especially deterministic rules that are interesting from a substantive point of view have been proposed (see Maris, t995; Van Mechelen, De Boeck,& Rosenberg, 1995; Maris. De Boeck,& Van Mechelen, 1996). For instance, according to a disjunctive communutality rule, the perception of a certain feature in both elements (object and attribute) is a sufficient condition for the elements to be associated, whereas, with a conju.nctive dotminance rule it is necessary that all the features that are perceived in one element (object) are also perceived in the other element (attribute). To make this abstract description of the PMD mo)del mnore specific it suffices to elaborate on the nature of the latent features and the mapping rule in some of the previously mentioned applications. In the application on psychiatric diagnosis, the latent fealures were conceived as psychiatric syndromes, This means that clinicians, wheni judging whether a patient has a symptom, covertly classify the patient according to which syndromes he or she has and also covertly classify the symptom as to which syndrornes it belongs to. Furthermore, a disjunctive mapping a le was proposed which meanis that a patient has a particular symptom according to a clinician if the patient has at least one syndrome which also contains that symnptomn. In the application on the facial perception of emiotions, the latent features represent properties of the face (e.g., the eyes of a fearful face, the mouth of a happy face). Applying a PMD model to these dlata, it is assumed that, when judging whether they perceive a certain emotion in a certain facial expression, raters may (or mlay not) perceive each of a set of relevant latent features in the face. In addition, it is assumed that in order to be perceived in a facial expression, emotions may require the perception of certain features. Furthermore, a conjunctive rule was used which means that an emotion is only perceived in a face if all the features that are required by the emotion are also perceived in the face.

When using the P'MD model for statistical intference, the following three problems need to be solved: (a) paramieter estimation, (b) imodel identifiability and (c) checking the model's goodness of fit. For the first problem, Maris et al. (1996) describe an EM algorithnm to obtain maximum likelihlood estimates and posterior mode estimates of the parameters of the PMD model, making use of the fact that the distribution of the augmented data has a simple structure. A method to estimate the standard errors of the parameters, however, is not presented by these authors. The second problem, model identifiability, is still to be investigated. It concerns the existence of trade-off relations in the model's parameters as well as the 155

Bayesian inference with probability matrix decomposition models

上一页第3页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email