Abstract
A common strategy for the analysis of object-attribute associations is to derive a low- dimensional spatial representation of objects and attributes which involves a compensatory model (e.g., principal components analysis) to explain the strength of object-attribute associations. As an alternative, probabilistic latent feature models assume that objects and attributes can be represented as a set of binary latent features and that the strength of object-attribute associations can be explained as a non-compensatory (e.g., disjunctive or conjunctive) mapping of latent features. In this paper, we describe the R package plfm which comprises functions for conducting both classical and Bayesian probabilistic latent feature analysis with disjunctive or a conjunctive mapping rules. Print and summary functions are included to summarize results on parameter estimation, model selection and the goodness of fit of the models. As an example the functions of plfm are used to analyze product-attribute data on the perception of car models, and situation-behavior associations on the situational determinants of anger-related behavior.
Highlights
The analysis of a two-way frequency table is a basic task in data analysis which is of interest to researchers in various domains of applied research (Agresti 2002)
A multinomial distribution is appropriate for modelling the frequencies in a two-way contingency table, and a Poisson distribution is appropriate for modelling frequencies which represent counts. Another type of two-way frequency data, which is the focus of the present paper, arises when the frequencies are derived by aggregating three-way three-mode or two-way three-mode binary observations
The aim of this paper is to present the R (R Core Team 2013) package plfm (Meulders 2013) for analyzing two-way frequency data with the non-compensatory probabilistic latent feature models (PLFMs) which were originally introduced by Maris et al (1996)
Summary
The analysis of a two-way frequency table is a basic task in data analysis which is of interest to researchers in various domains of applied research (Agresti 2002). The PLFM is related to the above described dimension-reduction techniques (PCA, CA) and to EFA in that it aims to explain the observed frequencies by representing row- and column elements in terms of a small set of latent variables It differs from the dimension-based approaches in that it represents row- and column elements in terms of binary latent features (instead of continuous dimensions), and in that it explains observed associations as a non-compensatory (i.e., disjunctive or conjunctive) function of feature patterns (rather than as a compensatory function).
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
