Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Latent Variables in Socio-Economic Models

Latent Variables in Socio-Economic Models 118 Reviews [part I, question of luck. Many potentially interesting research projects, as we all know, end up in dead ends, whereas apparently dull and routine problems can turn out to have unexpectedly interesting aspects. In both cases a degree of maturity is required in judging the potential worth of a projected piece of research which it would be unfair to expect the average student to possess. As this book well illustrates, the project technique can be an excellent vehicle for teaching and learning for many (perhaps most) students, but in my experience not all. As a means of assessment of achievement, however, it raises problems which teachers of economics are only just beginning to become aware of. R. K. WILKINSON Sheffield University 14. Latent Variables in Socio-economic Models. Edited by D. J. Aigner and A. S. Goldberger. Amsterdam, North-Holland, 1977. xiv, 383 p. 23 em. $48.95. In the last few years there has been a strong resurgence of interest in "latent variable models" among those concerned with the analysis of econometric, psychometric and sociological data. The term refers to models in which some of the variables of interest are not observable. More specifically, the variables which enter the "structural model" may not be observable; one has also the "observational model" which describes how observables are statistically related to these variables (are "multiple indicators" of them, in the jargon). "Regression with both variables subject to error" is an early example of the genre. Factor analysis and Sewell Wright's technique of path coefficients are two other examples of relatively early attempts to formulate and develop inference procedures for latent variable models. The incomprehension and suspicion with which these methods were viewed over many years is a measure of the degree of sophistication needed to formulate such models, to see them as representing real situations and to develop a theory of inference for them. The new wave of interest which stimulated the articles collected in this volume establishes the subject of latent variable models as respectable, well-defined and practically important. The list of contributors (Aigner, Bielby, Browne, Chamberlain, Featherman, Ferrara, Geraci, Geweke, Goldberger, Griliches, Hauser, Joreskog, Kadane, Maravall, McCallum, McGuire, Mouchart, Robinson, Sanday, Sorbom, Staelin, Zellner) includes several pioneers of this second wave. The papers by Goldberger, Zellner, Griliches and Robinson have been particularly significant in re-awakening interest. The two later papers by Joreskog and Joreskog/Sorbom are quite plainly those in which the model finds a clear, general, compact formulation and statistical analysis. They include two acronyms which have probably come to stay: ACOVS (analysis of covariance structures) and LISREL (linear structural relations). The volume includes none of the extensive work by Wold, which has been somewhat divorced from this literature, but which precedes most of it as a recognition of the theoretical and practical importance of latent variable models. Topics covered in the volume include the central questions of identifiability and of properties of various estimators (maximum likelihood, generalized, least squares, etc.). Simultaneity, and the contribution of instrumental variables, are also discussed. One good point is that certain models have a higher degree of identifiability in dynamic than in static versions. The papers are collected from the literature, but have been re-set in a uniform type, with slight re-editing in some cases. As far as I can see, no cross-referencing has been provided additional to references made in the original papers. In reading such a collection of papers, one is un­ reasonably irritated by apparently seeing the one problem, in variants which are often only slight, repeatedly described and discussed without apparent major advance along the main line-history resembles a bad dream in this respect. However, to react thus to history is unreasonable, and advance is recognizable both statistically over time, and in occasional leaps. The two Joreskog papers mentioned are welcome just because of their clean, powerful, summary character. The volume is obviously of great value and interest to anybody concerned with multivariate techniques, or with models for the social sciences. The subject certainly needs an organized text. However, this volume makes clear what such a text probably would not: that the running in the development of statistical methodology is often made by non-statisticians. Or, to put it more fairly: that methodology develops from applications rather than from theory. P. WHITTLE Cambridge University http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the Royal Statistical Society Series A (Statistics in Society) Oxford University Press

Latent Variables in Socio-Economic Models

Loading next page...
 
/lp/oxford-university-press/latent-variables-in-socio-economic-models-PGtcE6fn2m

References (0)

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Copyright
© 1978 The Authors
ISSN
0964-1998
eISSN
1467-985X
DOI
10.2307/2344794
Publisher site
See Article on Publisher Site

Abstract

118 Reviews [part I, question of luck. Many potentially interesting research projects, as we all know, end up in dead ends, whereas apparently dull and routine problems can turn out to have unexpectedly interesting aspects. In both cases a degree of maturity is required in judging the potential worth of a projected piece of research which it would be unfair to expect the average student to possess. As this book well illustrates, the project technique can be an excellent vehicle for teaching and learning for many (perhaps most) students, but in my experience not all. As a means of assessment of achievement, however, it raises problems which teachers of economics are only just beginning to become aware of. R. K. WILKINSON Sheffield University 14. Latent Variables in Socio-economic Models. Edited by D. J. Aigner and A. S. Goldberger. Amsterdam, North-Holland, 1977. xiv, 383 p. 23 em. $48.95. In the last few years there has been a strong resurgence of interest in "latent variable models" among those concerned with the analysis of econometric, psychometric and sociological data. The term refers to models in which some of the variables of interest are not observable. More specifically, the variables which enter the "structural model" may not be observable; one has also the "observational model" which describes how observables are statistically related to these variables (are "multiple indicators" of them, in the jargon). "Regression with both variables subject to error" is an early example of the genre. Factor analysis and Sewell Wright's technique of path coefficients are two other examples of relatively early attempts to formulate and develop inference procedures for latent variable models. The incomprehension and suspicion with which these methods were viewed over many years is a measure of the degree of sophistication needed to formulate such models, to see them as representing real situations and to develop a theory of inference for them. The new wave of interest which stimulated the articles collected in this volume establishes the subject of latent variable models as respectable, well-defined and practically important. The list of contributors (Aigner, Bielby, Browne, Chamberlain, Featherman, Ferrara, Geraci, Geweke, Goldberger, Griliches, Hauser, Joreskog, Kadane, Maravall, McCallum, McGuire, Mouchart, Robinson, Sanday, Sorbom, Staelin, Zellner) includes several pioneers of this second wave. The papers by Goldberger, Zellner, Griliches and Robinson have been particularly significant in re-awakening interest. The two later papers by Joreskog and Joreskog/Sorbom are quite plainly those in which the model finds a clear, general, compact formulation and statistical analysis. They include two acronyms which have probably come to stay: ACOVS (analysis of covariance structures) and LISREL (linear structural relations). The volume includes none of the extensive work by Wold, which has been somewhat divorced from this literature, but which precedes most of it as a recognition of the theoretical and practical importance of latent variable models. Topics covered in the volume include the central questions of identifiability and of properties of various estimators (maximum likelihood, generalized, least squares, etc.). Simultaneity, and the contribution of instrumental variables, are also discussed. One good point is that certain models have a higher degree of identifiability in dynamic than in static versions. The papers are collected from the literature, but have been re-set in a uniform type, with slight re-editing in some cases. As far as I can see, no cross-referencing has been provided additional to references made in the original papers. In reading such a collection of papers, one is un­ reasonably irritated by apparently seeing the one problem, in variants which are often only slight, repeatedly described and discussed without apparent major advance along the main line-history resembles a bad dream in this respect. However, to react thus to history is unreasonable, and advance is recognizable both statistically over time, and in occasional leaps. The two Joreskog papers mentioned are welcome just because of their clean, powerful, summary character. The volume is obviously of great value and interest to anybody concerned with multivariate techniques, or with models for the social sciences. The subject certainly needs an organized text. However, this volume makes clear what such a text probably would not: that the running in the development of statistical methodology is often made by non-statisticians. Or, to put it more fairly: that methodology develops from applications rather than from theory. P. WHITTLE Cambridge University

Journal

Journal of the Royal Statistical Society Series A (Statistics in Society)Oxford University Press

Published: Dec 5, 2018

There are no references for this article.