Scaling

A Sourcebook for Behavioral Scientists

Gary Maranell

• Publisher: Routledge
• Year: 2017
• ISBN:
  9781351492065
• Edition: 1
• Topic: Mathematics
• Subtopic: Applied Mathematics
• Index: Mathematics

Despite the obvious importance of measurement in any scientific endeavor, few students of the social sciences receive adequate training in the principles and problems of assigning numerical values to the subjects, objects, events, groups and operations they study, and still less in the process of translating theoretical ideas and concepts into variables. This kind of casualness with respect to measurement is often in marked contrast to their methodically designed research, which has grown out of subtle and sophisticated theoretical consideration.Scaling is intended to remedy this deficiency by providing a broad and detailed description of the major processes for developing measurement scales. The chapters, which include both classics in the field and the best of modern work, require no great mathematical sophistication, and go well beyond the conventional study of attitudes to the more general uses of scaling. They enable the student and researcher to examine the development of measures of scalability and the problems and weaknesses they present, to become familiar with the development of tests of significance for reproducibility and scalability and the need for them, and to examine the lively history of the subject and experience the excitement that can be secured from sharing with a creative author the first report of his insight.Part One presents a series of general articles that deal in philosophic terms with the problem of measurement, with what is meant by measurement and scaling as well as the notions underlying the process of measuring. Part Two deals with the scaling methods developed by L. L. Thurstone, including paired comparison scaling, equal-appearing interval scaling, and successive interval scaling. The third part focuses upon scalogram analysis, presenting the background, rationale and procedures for Guttman scaling. The fourth part is concerned with summated rating, or Likert scaling. Part Five is a consideration of unfold

Introduction

This book focuses upon the process of scaling as a problem of empirical measurement and, even more importantly, as the process of translating theoretical ideas and concepts into variables. These variables or systems of observations, which are coded in a sufficiently reliable and precise manner, are necessary to make it possible for scientists to perceive, record, and study differences, correlations, and change in the phenomena they are interested in understanding. Without the scaling procedures that record and measure variables, the process of studying social and individual behavior would be seriously limited, and research in many areas of study would be made impossible. Behavioral science may be said to have advanced in many respects with the developments that have occurred in the level of sophistication of scaling and measurement procedures. Such continued advances, coupled with an increasing sophistication in theory and model construction, are prerequisites for its subsequent development.

Discussions of measurement often advance the accurate but rather proasaic assertion that “measurement involves the assignment of numbers, or sometimes numerals, to objects according to rules.” This volume attempts to provide a variety of such sets of rules. Sections II through V each describe a different procedure, or in some cases a set of procedures, for the assignment of numbers to objects. These procedures are the operations needed to construct the measuring devices that behavioral scientists use to do the research and theory testing necessary to develop and advance their field of study. In a fundamental sense, the procedures of scaling are consonant with and subtlely and consistently encouraging of more systematic and precise problem formation and theory construction. This is because scaling and measurement procedures strongly encourage and, in most instances, demand an attention to and a focusing upon variables that are quantifiable rather than upon less precise pieces of observation of social and psychological phenomena. In an important sense, one fundamental prerequisite for the continued development of the behavioral sciences is the increasing quantification of our basic variables that, in turn, demands some theory and technique of measurement.

This volume is also an attempt to provide a glimpse of the growth and development of scaling as it has been employed in the behavioral sciences. The development of scaling methods has been somewhat cumulative, each approach building in some ways upon antecedent attacks upon the measurement problem. In the instances when this development has not been cumulative, the innovation was often stimulated by a shortcoming, an unrealistic or unacceptable assumption, or a weakness in the available methods. In some of these cases, the innovation has been based upon alternative assumptions that seem more realistic or are easier to accept. The development within scaling also involves an elaboration and improvement of techniques and procedures and a needed attention to oversights and to the dangers involved in the misapplication and the naive use of particular scaling methods.

Many behavioral scientists have been somewhat indifferent to the problems and issues involved in measurement and scaling. This frequently has led them to allocate much less thought to this aspect of their research than is necessary for careful inquiry. When measurement problems are ignored, the effects and the results are not unlike those that would be encountered by astronomers who are forced to use cracked lens and no calibration or surveyers who are forced to use rubber yardsticks or no yardsticks at all, or physicists with, watches which run randomly fast and slow. In view of this one can only be dismayed at the consistently cavalier approach to measurement which has frequently typified behavioral science research. Superficial and naive measurements are occasionally subjected to inappropriately careful analysis that, of course, is not better or worse than the instances when the grossness of the analysis rivals the casualness of the measurement. This casualness in regard to measurement is often in marked contrast to the fact that such gross measures are used to do research that has been otherwise methodically designed and has grown out of extremely subtle, careful, lucid, and sophisticated theoretical consideration. It is safe to say that an analysis is nearly meaningless when the phenomena being analyzed have been carelessly measured.

The indifference to measurement, or its lack of a position of importance in many of the behavioral sciences (particularly Sociology, Political Science, and Anthropology), suggests that it is seen as peripheral rather than centrally important to the field. This lack of integration of something as important as measurement into the body of the behavioral sciences is a matter of some seriousness. The fact of this lack of integration is revealed through an examination of course offerings in many behavioral science departments. These listings of courses reveal that there exists an apparent plethora of courses on all facets of, for example, the Sociology of----, and that independent from all these courses there is a set of courses on how to do the science—in case anyone might be interested. These separate and peripheral courses include those devoted to scaling and measurement, design and analysis and are taught by a species, unique to the less established sciences, called methodologists. Apparently few are concerned about how a science can develop when it separates training in the ways of doing the science from knowledge of the content of the science.

There are many reasons why scaling and measurement are important to the behavioral sciences. First, science is nomothetic, or put another way, nomothetic measurement is necessary to do scientific research and is needed to develop a science. This means that science finds it necessary to measure or scale the same property in all the units, subjects, or phenomena being studied. The relationship between two variables can be discovered only when all subjects or units of study have been scaled or measured on the same two variables. Obviously the nomothetic nature of science strongly encourages the use of scales of measurement among behavioral scientists. The construction of a scale designed to measure, for example, anxiety does, when it is applied or administered to all the members of a group of subjects, provide nomothetic measurement. Similarly the employment of a scale of urban integration to measure this property in a sample of cities provides an assessment of the same property in all the cities being studied.

A second reason scaling and measurement are important to the development of science is that they provide standardization. For a science to grow other researchers must be able to replicate one’s research, and to do this others must be able to perform the same measurements one has previously performed. This means that the procedures of measurement must be clear, open or public, and repeatable, and the easiest way to accomplish this is through the development, use, and publication of the scale or measurement device. When one has conducted a study and has employed a scale of, for example, social integration or cohesion, other researchers can by employing the same scales secure comparable measurement in other samples, for other studies and purposes. A related advantage is that by being standardized a scale increases objectivity since the biases of each researcher are controlled and minimized by using the standardized measurement. Furthermore even the biases of the scientist that devised the scale are, by virtue of the openness and availablity of the scale, subject to greater scrutiny and consequently greater control.

Another advantage of scaling is that it provides for increased precision. Instead of simply classifying groups in terms of being formalized or not, a scale that identifies the characteristics and indicators of the variable “formalization” allows the scientist to make more extensive and subtle discriminations between groups on this variable and permits a more precise description of the ways the groups vary in formalization.

A fourth advantage of scaling and measurement is that they allow for and increase conciseness. Quantitative measurement and scaling permit the condensation and conciseness of mathematics to be employed by scientist. This conciseness occurs in the description of the quantitative measurement, the reporting of the results of the measurement, as well as in the description of the theoretical and empirical relationships discovered or predicted. This advantage of scaling and measurement involves more explicitly an increased economy. The economy is involved in the time, effort, and words gained or saved when one expresses theories, variables, relationships, and findings in general, as well as in the processes involved in quantitative and mathematical analysis. It is for reasons such as these that some have said, “If you haven’t measured it, you do not know what you are talking about.”

The present limited concern for scaling and scale construction grew out of a need to secure useful measurement of phenomena that were of interest to behavioral scientists and did not lend themselves easily to quantification—that is, the development of scaling was stimulated by an interest in securing something Thurstone referred to as “subjective measurement.” Behavioral scientists were and still are interested in obtaining measurements of phenomena such as values, attitudes, and opinions, in addition to other phenomena. In the 1920’s, Thurstone, having these interests, considered the issues involved in measurement seriously and provided the early and still substantial developments associated with his name. His substantive empirical research illustrated the usefulness of the techniques he developed. Thurstone studied food preferences; social values; views of criminal offenses; attitudes toward war, the church, and movies among other things. The continued use of these methods documents the importance of his accomplishments.

Summated rating scales have also focused upon measuring subjective phenomena, explicitly attitudes and opinions. Because of Rensis Likert’s advocacy of the method, as well as his contributions to the method, such scales are sometimes referred to as Likert scales. He viewed the process as an indirect method of measuring dispositions and values that are most easily expressed verbally. Scales of this type, which involve the summing of weighted responses, have several advantages. The advantages include the facts that such scales are easy to construct, have been found to be highly reliable, and have consistently provided useful and meaningful results in research. Therefore, it is not surprising to find that Likert scaling is still very commonly used.

Louis Guttman’s interest was in the area of attitude measurement as well. Guttman scaling theory developed from the basic concept of a scalable or unidimensional universe of attributes. One of the main purposes of the technique is to ascertain whether the attitudes or characteristics (universe of content or attributes) being studied actually involve only a single dimension, or form a unidimensional scale. Guttman’s association with the research group that produced The American Soldier (Stouffer, et al., 1950) provided an opportunity for an appreciation of the importance of his contribution to disseminate quite widely. The publication also explained the theoretical background, procedures, and modes of evaluation of the cumulative or Guttman scales in a comprehensive manner and reported the results of a program of research that constructed and employed hundreds of such scales, measuring attitude areas such as soldiers’ adjustment to military life, the effects of combat, and postwar plans. The interest in cumulative scales has remained high, and it is still one of the most commonly employed methods of scaling many sociological, psychological, and political variables.

A more recent innovative contribution is that of C. H. Coombs. This method, called unfolding, is also used for attitude scaling although it is useful for scaling other objects and subject preferences as well. The procedure derives information on the unidimensionality and relative spacing o...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Preface
• Introduction
• Part I. Measurement and Scaling
• Part II. Thurstonian Methods
• Part III. Scalogram Analysis
• Part IV. Summated Rating Methods
• Part V. Unfolding Theory
• Part VI. Related Materials
• Selected Bibliography
• Appendix: Tables
• Index