 | |
| Author | L. L. Thurstone |
|---|---|
| Country | United States |
| Language | English |
| Subjects | Factor analysis, psychometrics |
| Publisher | University of Chicago Press |
Publication date | August 1935 |
| Media type | |
| Pages | 266 |
The Vectors of Mind [1] is a book published by American psychologist Louis Leon Thurstone in 1935 that summarized Thurstone's methodology for multiple factor analysis. [2]
The Vectors of Mind presents Thurstone's methods for conducting a factor analysis on a set of variables that allow for more than one factor, an important extension of Spearman's unifactor method. Having multiple factors adds significant complications and much of the book is focussed on the problem of rotation. It attempts to solve this problem by providing an objective basis for the rotation factors, called simple structure, and advocates the use of oblique (correlated) factors to achieve a simple structure. The book utilizes his centroid method of factor extraction, which made it feasible to complete the arduous calculations necessary for a factor analysis at a time when fast electronic computers had not even been imagined. This is a predominantly technical book that relies heavily upon mathematical presentations and provides multiple numerical examples. However, the early chapters delve into philosophical questions of the nature of science and present Thurstone's understanding of measurement theory.
Preface. This book extends and presents more formally the findings from the author's Multiple Factor Analysis paper of 1931. The author notes that he only recently learned matrix theory and presumes that other psychologists have had similar limitations in their training. He finds existing textbooks on the topic inadequate and the book begins with a presentation of matrix theory, written for those with undergraduate instruction in analytic geometry and real number calculus. The author expresses indebted to various professors in the mathematics department of the University of Chicago for helping him to develop his ideas. He also expresses appreciation to his computer (a person, Leone Chesire), who also wrote the appendix on the calculations used in the centroid method. He foresees a bright future for the use of factor analysis and expects to see the simplification of the computational methods. He expects factor analysis to become an important technique int the early stages of science. For example, the laws of classical mechanics could have been revealed by a factor analysis, by analyzing a great many attributes of objects that are dropped or thrown from an elevated point, with the time of fall factor uncorrelated with the weight factor. Work by Sewell Wright on path coefficients and Truman L. Kelley on multiple factors differs from factor analysis, which Thurstone sees as an extension of professor Spearman's work.
Mathematical Introduction. A brief presentation of matrices, determinants, matrix multiplication, diagonal matrices, the inverse, the characteristic equation, summation notation, linear dependence, geometric interpretations, orthogonal transformations, and oblique transformations.
Chapter I. The Factor Problem. Natural phenomena are only comprehensible through constructs that are man-made inventions. A scientific law is not part of nature; it is but man's way of understanding nature. Examples are provided of such man-made constructs from physics. He responds to skepticism from the practitioners of "rigorous science" that human behavior can ever be brought into the fold of such science by pointing out that there is considerable individuality in physical events even though described by rigorous scientific laws, such as the fact that every explosion is unique. Human abilities are the cause of individual differences in the "completion of a task". The science of psychology will reduce a large number of psychological abilities down to primary reference traits. Formal definitions are provided for the concepts of trait, ability, test, score, linear independence, statistical independence, experimental independence, reference abilities, primary abilities, and unitary ability. These conceptions constitute a theory of measurement that defines factors common to all tests in a battery-the communality of the test battery-, a specific factor that is unique to one test-the specificity of the test-, and the error variance. Factor analysis can determine the communality of a test, but cannot separate the uniqueness into the specific factor and the error factors. The reliability coefficient is the sum of communality and the specificity of a test.
Chapter II. The Fundamental Factor Theorem. The factor matrix post-multiplied by its transpose gives the reduced correlation matrix: this is the fundamental factor theorem. The task of factor analysis is to find a factor matrix of the lowest possible rank (the least number of factors) that can reproduce the off-diagonal members of the observed correlation matrix as close as can be expected, allowing for sample variation. The bulk of the chapter considers mathematical issues, including the rank of a matrix and methods for estimating the commonalities of the correlation matrix (the diagonal elements).
Chapter III: The Centroid Method. A computation method is developed for factoring a correlation matrix, which is a symmetric matrix of real elements. After a conceptual presentation of the method, some worked examples are provided, including one with eight variables, and another with fifteen variables that are factored into four factors. The mechanics of the calculations are given in Appendix I, which provides the specific steps in making the calculation (the algorithm).
Chapter IV: The Principal Axes. A method is presented for determining a desirable rotation of the orthogonal factors called the principal axes. The mathematical foundations are provided, as well as worked examples. This approach is distinguished from Hotelling's method, which the author feels has limited usefulness to factor analysis. The unrotated solution for 15 psychological tests given in chapter III are rotated to their principal axes.
Chapter V: The Special Case of Rank One. Spearman presented factor analysis with a single factor (a matrix with rank one) thirty years, but recent advances have made it possible to extend factor analysis to multiple factors. The shortcomings of Spearman's method of tetrad differences are detailed and the current approach found to be more accurate. A numerical example is given.
Chapter VI: Primary Traits. Rotation does not affect the results of the fundamental factor theorem. All rotations result in the same reduced correlation matrix so other criteria must be used to ascertain the best rotation. Criteria refer to "simple structure": the book presents very detailed criteria for simple structure, but more generally it consists of minimizing the number of loading for each variable and wide variance for loadings of each factor. Realizing simple structure may require uses of oblique (correlated) factors. Three additional criteria are given that define when the simple structure is unique. Graphical-mathematical methods are developed for understanding and defining the structure that reveals primary traits–the scientific goal of factor analysis. The prior worked example of fifteen psychological traits is rotated to an oblique simple structure to reveal three intercorrelated primary traits.
Chapters VII - X: The remaining chapters explore more specific details and problems that can arise. Chapter VII considers several methods for isolating primary traits, with numerical examples given. Chapter VIII addresses the methodological problems that can arise when the correlation matrix has negative correlations. Though most scientific investigations of primary abilities will entail oblique factors, there are situations where the factors are likely to be orthogonal. Chapter IX looks at techniques for achieving orthogonal rotations. The results of a factor analysis can be used to estimate each individual's score on the primary abilities based upon the individual's scores on the tests. Chapter X presents a method for obtaining the regression weights for estimating primary abilities from subject scores, and well as for estimating subjects scores from the primary traits (for estimating the components of variance of the subject scores).
Appendices. I: Outline of Calculations for the Centroid Method with Unknown Diagonals. II: A Method of Finding the Roots of a Polynomial. III: A Method of Determining the Square Root on the Calculating Machine.
In 1904 Charles Spearman published a paper that largely founded the field of psychometrics and included a crude form of factor analysis that attempted to determine if a single factor model was appropriate. [3] There was limited subsequent work on factor analysis until Thurstone published a paper in 1931 called Multiple Factor Analysis, [4] which expanded Spearman's single-factor analysis to include more than one factor. In 1932, Hotelling presented a more accurate method of extracting factors, which he called principal components analysis. [5] Thurstone rejected Hotelling's approach because it set the commonalities to 1.0, and Thurstone realized that will introduce distortions to the factor loadings when variables include unique components. Hotelling's method was also limited by the fact that it required too much calculation to be useable with more than about ten variables. [6] A year after Hotelling's paper, Thurstone presented a more efficient way of extracting factors, called the centroid method, [7] which allowed the factor analysis of a far larger number of variables. Later that year he gave his presidential address to the American Psychological Association wherein he presented the results of several factor analyses, including a factor analysis of 60 adjectives describing personality traits, showing how they could be reduced to five personality traits. He also presented analyses of 37 mental health symptoms, of attitudes towards 12 controversial social issues, and of 9 IQ tests. [8] In those analyses, Thurstone had made use of tetrachoric correlation coefficients, a method for estimating continuous variable correlations from dichotomous variables. Tetrachorics require extensive calculations but in early 1933, he and two colleagues at the University of Chicago published a set of computing diagrams that greatly reduce the calculations needed for those coefficients, [9] another aspect of making his method of factor analysis practical with more than just a few variables. His 1933 presidential address was published in early 1934 with the title Vectors of the Mind. It lacked methodological and mathematical details of his technique, which is then the subject of this book. A 2004 conference called Factor Analysis at 100 produced a book with two chapters that document the historical importance Thurstone's contributions to factor analysis. [10] [11] Thurstone's approach to factor analysis remains an important method in psychological research and it has since been used in numerous other fields of study. [12] It is now considered part of a family of methods for analyzing the covariance structure of variables, which includes principal components analysis, exploratory factor analysis, confirmatory factor analysis, and structural equation modeling. [13]
Psychological statistics is application of formulas, theorems, numbers and laws to psychology. Statistical methods for psychology include development and application statistical theory and methods for modeling psychological data. These methods include psychometrics, factor analysis, experimental designs, and Bayesian statistics. The article also discusses journals in the same field.
Psychometrics is a field of study within psychology concerned with the theory and technique of measurement. Psychometrics generally refers to specialized fields within psychology and education devoted to testing, measurement, assessment, and related activities. Psychometrics is concerned with the objective measurement of latent constructs that cannot be directly observed. Examples of latent constructs include intelligence, introversion, mental disorders, and educational achievement. The levels of individuals on nonobservable latent variables are inferred through mathematical modeling based on what is observed from individuals' responses to items on tests and scales.
Principal component analysis (PCA) is a popular technique for analyzing large datasets containing a high number of dimensions/features per observation, increasing the interpretability of data while preserving the maximum amount of information, and enabling the visualization of multidimensional data. Formally, PCA is a statistical technique for reducing the dimensionality of a dataset. This is accomplished by linearly transforming the data into a new coordinate system where the variation in the data can be described with fewer dimensions than the initial data. Many studies use the first two principal components in order to plot the data in two dimensions and to visually identify clusters of closely related data points. Principal component analysis has applications in many fields such as Population Genetics, Microbiome studies, Atmospheric Science etc.
In statistics, correlation or dependence is any statistical relationship, whether causal or not, between two random variables or bivariate data. Although in the broadest sense, "correlation" may indicate any type of association, in statistics it normally refers to the degree to which a pair of variables are linearly related. Familiar examples of dependent phenomena include the correlation between the height of parents and their offspring, and the correlation between the price of a good and the quantity the consumers are willing to purchase, as it is depicted in the so-called demand curve.
In statistics, Spearman's rank correlation coefficient or Spearman's ρ, named after Charles Spearman and often denoted by the Greek letter (rho) or as , is a nonparametric measure of rank correlation. It assesses how well the relationship between two variables can be described using a monotonic function.
Factor analysis is a statistical method used to describe variability among observed, correlated variables in terms of a potentially lower number of unobserved variables called factors. For example, it is possible that variations in six observed variables mainly reflect the variations in two unobserved (underlying) variables. Factor analysis searches for such joint variations in response to unobserved latent variables. The observed variables are modelled as linear combinations of the potential factors plus "error" terms, hence factor analysis can be thought of as a special case of errors-in-variables models.
In the social sciences, scaling is the process of measuring or ordering entities with respect to quantitative attributes or traits. For example, a scaling technique might involve estimating individuals' levels of extraversion, or the perceived quality of products. Certain methods of scaling permit estimation of magnitudes on a continuum, while other methods provide only for relative ordering of the entities.
The g factor is a construct developed in psychometric investigations of cognitive abilities and human intelligence. It is a variable that summarizes positive correlations among different cognitive tasks, reflecting the fact that an individual's performance on one type of cognitive task tends to be comparable to that person's performance on other kinds of cognitive tasks. The g factor typically accounts for 40 to 50 percent of the between-individual performance differences on a given cognitive test, and composite scores based on many tests are frequently regarded as estimates of individuals' standing on the g factor. The terms IQ, general intelligence, general cognitive ability, general mental ability, and simply intelligence are often used interchangeably to refer to this common core shared by cognitive tests. However, the g factor itself is merely a mathematical construct indicating the level of observed correlation between cognitive tasks. The measured value of this construct depends on the cognitive tasks that are used, and little is known about the underlying causes of the observed correlations.
Louis Leon Thurstone was an American pioneer in the fields of psychometrics and psychophysics. He conceived the approach to measurement known as the law of comparative judgment, and is well known for his contributions to factor analysis. A Review of General Psychology survey, published in 2002, ranked Thurstone as the 88th most cited psychologist of the 20th century, tied with John Garcia, James J. Gibson, David Rumelhart, Margaret Floy Washburn, and Robert S. Woodworth.
Raymond Bernard Cattell was a British-American psychologist, known for his psychometric research into intrapersonal psychological structure. His work also explored the basic dimensions of personality and temperament, the range of cognitive abilities, the dynamic dimensions of motivation and emotion, the clinical dimensions of abnormal personality, patterns of group syntality and social behavior, applications of personality research to psychotherapy and learning theory, predictors of creativity and achievement, and many multivariate research methods including the refinement of factor analytic methods for exploring and measuring these domains. Cattell authored, co-authored, or edited almost 60 scholarly books, more than 500 research articles, and over 30 standardized psychometric tests, questionnaires, and rating scales. According to a widely cited ranking, Cattell was the 16th most eminent, 7th most cited in the scientific journal literature, and among the most productive psychologists of the 20th century. He was, however, a controversial figure, due in part to his friendships with and intellectual respect for white supremacists and neo-Nazis.
Charles Edward Spearman, FRS was an English psychologist known for work in statistics, as a pioneer of factor analysis, and for Spearman's rank correlation coefficient. He also did seminal work on models for human intelligence, including his theory that disparate cognitive test scores reflect a single General intelligence factor and coining the term g factor.
In statistics, canonical analysis belongs to the family of regression methods for data analysis. Regression analysis quantifies a relationship between a predictor variable and a criterion variable by the coefficient of correlation r, coefficient of determination r2, and the standard regression coefficient β. Multiple regression analysis expresses a relationship between a set of predictor variables and a single criterion variable by the multiple correlation R, multiple coefficient of determination R², and a set of standard partial regression weights β1, β2, etc. Canonical variate analysis captures a relationship between a set of predictor variables and a set of criterion variables by the canonical correlations ρ1, ρ2, ..., and by the sets of canonical weights C and D.
In statistics, confirmatory factor analysis (CFA) is a special form of factor analysis, most commonly used in social research. It is used to test whether measures of a construct are consistent with a researcher's understanding of the nature of that construct. As such, the objective of confirmatory factor analysis is to test whether the data fit a hypothesized measurement model. This hypothesized model is based on theory and/or previous analytic research. CFA was first developed by Jöreskog (1969) and has built upon and replaced older methods of analyzing construct validity such as the MTMM Matrix as described in Campbell & Fiske (1959).
Peter Hans Schönemann was a German born psychometrician and statistical expert. He was professor emeritus in the Department of Psychological Sciences at Purdue University. His research interests included multivariate statistics, multidimensional scaling and measurement, quantitative behavior genetics, test theory and mathematical tools for social scientists. He published around 90 papers dealing mainly with the subjects of psychometrics and mathematical scaling. Schönemann's influences included Louis Guttman, Lee Cronbach, Oscar Kempthorne and Henry Kaiser.
Psychometrika is the official journal of the Psychometric Society, a professional body devoted to psychometrics and quantitative psychology. The journal covers quantitative methods for measurement and evaluation of human behavior, including statistical methods and other mathematical techniques. Past editors include Marion Richardson, Dorothy Adkins, Norman Cliff, and Willem J. Heiser. According to Journal Citation Reports, the journal had a 2019 impact factor of 1.959.
In multivariate statistics, exploratory factor analysis (EFA) is a statistical method used to uncover the underlying structure of a relatively large set of variables. EFA is a technique within factor analysis whose overarching goal is to identify the underlying relationships between measured variables. It is commonly used by researchers when developing a scale and serves to identify a set of latent constructs underlying a battery of measured variables. It should be used when the researcher has no a priori hypothesis about factors or patterns of measured variables. Measured variables are any one of several attributes of people that may be observed and measured. Examples of measured variables could be the physical height, weight, and pulse rate of a human being. Usually, researchers would have a large number of measured variables, which are assumed to be related to a smaller number of "unobserved" factors. Researchers must carefully consider the number of measured variables to include in the analysis. EFA procedures are more accurate when each factor is represented by multiple measured variables in the analysis.
Cultural consensus theory is an approach to information pooling which supports a framework for the measurement and evaluation of beliefs as cultural; shared to some extent by a group of individuals. Cultural consensus models guide the aggregation of responses from individuals to estimate (1) the culturally appropriate answers to a series of related questions and (2) individual competence in answering those questions. The theory is applicable when there is sufficient agreement across people to assume that a single set of answers exists. The agreement between pairs of individuals is used to estimate individual cultural competence. Answers are estimated by weighting responses of individuals by their competence and then combining responses.
In statistics, factor analysis of mixed data or factorial analysis of mixed data, is the factorial method devoted to data tables in which a group of individuals is described both by quantitative and qualitative variables. It belongs to the exploratory methods developed by the French school called Analyse des données founded by Jean-Paul Benzécri.
Charles Spearman developed his two-factor theory of intelligence using factor analysis. His research not only led him to develop the concept of the g factor of general intelligence, but also the s factor of specific intellectual abilities. L. L. Thurstone, Howard Gardner, and Robert Sternberg also researched the structure of intelligence, and in analyzing their data, concluded that a single underlying factor was influencing the general intelligence of individuals. However, Spearman was criticized in 1916 by Godfrey Thomson, who claimed that the evidence was not as crucial as it seemed. Modern research is still expanding this theory by investigating Spearman's law of diminishing returns, and adding connected concepts to the research.