Elements of Dual Scaling

An Introduction To Practical Data Analysis

Shizuhiko Nishisato

• Publisher: Psychology Press
• Year: 2014
• eBook ISBN: 9781317781936
• Topic: Psicología
• Subtopic: Historia y teoría en psicología

Quantification methodology of categorical data is a popular topic in many branches of science. Most books, however, are either too advanced for those who need it, or too elementary to gain insight into its potential. This book fills the gap between these extremes, and provides specialists with an easy and comprehensive reference, and others with a complete treatment of dual scaling methodology -- starting with motivating examples, followed by an introductory discussion of necessary quantitative skills, and ending with different perpsectives on dual scaling with examples, advanced topics, and future possibilities.

This book attempts to successively upgrade readers' readiness for handling analysis of qualitative, categorical, and non-metric data, without overloading them. The writing style is very friendly, and difficult topics are always accompanied by simple illlustrative examples.

There are a number of topics on dual scaling which were previously addressed only in journal articles or in publications that are not readily available. Integration of these topics into the standard framework makes the current book unique, and its extensive coverage of relevant topics is unprecedented. This book will serve as both reference and textbook for all those who want to analyze categorical data effectively.

PART I
Background

Part 1 contains six chapters to introduce basic ideas on data analysis, data types and dual scaling. If you are already familiar with dual scaling, correspondence analysis, Hayashi's theory of quantification, optimal scaling, or homogeneity analysis, you may skip Part 1, and start with Part 2. However, if you are not familiar with any one of these terms, try to digest the material covered here before you engage yourself in applications of dual scaling in Part 2. Those topics in Part 1 will prove important for subsequent chapters even if they may look elementary.

CHAPTER 1
To Begin With

There are a few things you might want to look at before you get involved in dual scaling. They are concerned with the nature of data analysis. As you will see, most points to be mentioned here are what your common sense would dictate. Nevertheless, putting those points in one place may be of some use as an introduction to data analysis.

Data collection is your first topic. If you are a teacher, you may give several tests to a group of students, and their scores on the tests will be your data. If your work is in marketing research, your data may consist of consumers' responses to a set of so-called multiple-choice questions plus their biographical information (e.g., gender, age, profession). If you are a clinical psychologist, patients' responses to an inkblot test may be your data. If you are a public relations officer of a company, a list of complaints from the public would constitute a data set. Whatever your task, you must collect “valid” information, valid in the sense that it is worth analyzing and can be analyzed. This is a very important point with many relevant problems, yet frequently tends to be overlooked or ignored completely.

First of all, you must consider the task of the respondents or subjects: for instance, to answer a set of questions, to compare the taste of Coca Cola and Pepsi, or to rank five candidates for a committee of three. Are these tasks simple and clear enough for your subjects? It is easy to assume, because you are familiar with the area under investigation, that your subjects would be able to answer all the questions, say about pollution problems, social welfare problems, or mandatory retirement issues. If you want to solicit reactions of people in your community to the government's recent tax reform proposal, make the questionnaire short and easy to answer.

Once you know what you want to find out from data analysis, you must collect data suitable for your purposes. If two classes of students are to be compared on their performance in mathematics, for example, make sure that the same mathematics test is given to both classes, and avoid such a day for testing when many students in one class are absent from school. Would you not also wish to discover what other information might be related to their differences in achievement, for instance, aptitude, career orientation, curricula, use of computers, and so on? If you do, collect the information from the students.

In many situations, you want to find out whatever your data can tell you. As compared with data collection for the purpose of research, it might be more difficult to collect data for so-called exploratory work, where you want to gather not only specific but also general information. In an attempt to collect as much information as possible, however, don't be too ambitious. There are many studies in which the investigator employs so-called open-ended questionnaires consisting of such a question as “What do you think of capital punishment? State your view in the following blank space.” There is nothing wrong with this kind of open-ended question if you want to collect opinions about an issue. However, from the data analysis point of view, you would definitely prefer hearing from ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Dedication
• Table of Contents
• Preface
• Part I Background
• Part II Incidence Data
• Part III Dominance Data
• Part IV Special Topics
• Appendix
• References
• Author Index
• Subject Index