Jean L. Freeman

Strategies in the Multivariate Analysis of Data from Complex Surveys

Summary Certain aspects of the multivariate analysis of data from possibly complex survey designs are discussed in terms of a large sample methodology involving weighted least squares algorithms for the computation of Wald statistics. In particular, consideration is given to the problems associated with univariate and multivariate comparisons among cross-classified sub-populations. Finally, a philosophical discussion based on response (in the sense of measurement) error model concepts is given to motivate the validity of such least squares (both unweighted and weighted) methods for investigating the relationship between a particular dependent variable and several independent variables in a manner analogous to multiple regression analysis. This methodology is illustrated with several examples involving data from reports of the U.S. National Center for Health Statistics.

Bail Risk: A Multivariate Analysis

B AIL, also called pretrial release, is a legal means of freeing a defendant before court disposition of criminal charges against him. Its purpose is to prevent the defendant from being jailed when still presumed innocent and to assure that he will appear in court when required. The right to bail is not absolute; the court may set reasonable conditions intended to insure the defendants appearance at the various stages of his trial. Failure to appear in court when required usually carries some penalty, which may take the form of forfeiting a specific sum of money, additional criminal punishment, or loss of pretrial freedom. The penalty reflects the risks that society takes when the defendant is released, including:

Strategies in the Multivariate Analysis of Data from Complex Surveys II: An Application to the United States National Health Interview Survey

Summary An implementation of a procedure for the analysis of complex sample surveys (Koch, Freeman and Freeman, 1975) is discussed. This study examines the effects of (1) post-sampling data adjustments (post-stratification), (2) pseudo-replication of rates for variance estimation and (3) the assumption of zero covariance among group estimates of physician visits per person per year. The first two effects are not important for this data set, but the estimation of inter-group or inter-domain covariances plays an important role in testing hypotheses and estimating variances of fitted models. In addition to the investigation of these effects a detailed discussion of the fitting of linear models for these procedures is given. In conclusion, the approach of Koch, Freeman and Freeman is found to be a reasonable and valid method of analysing and summarizing complex sample survey data.

An application of the clumped binomial model to the analysis of clustered attribute data

This paper is concerned with the statistical analysis of situations where the experimental units are clusters of subjects, each of whom is classified with respect to the presence of a particular (binary response) attribute. For this purpose, the clumped binomial probability model is used to define an underlying mean rate parameter in terms of which the effects of certain treatments as well as cluster size itself can be investigated. The resulting methodology involves a synthesis of certain maximum likelihood and weighted least squares procedures; it is illustrated for a three-dimensional contingency table pertaining to numbers of deaths in litters of baby mice.

The relationship between legal abortion and marriage

A study using the abortion-birth ratios for residents of each U.S. state for the second half of 1970 and for all of 1971 was done to determine if legal abortions reduced marriages. Data showed that trends were consistent with the hypothesis that a relationship exists between l egalized abortion-birth ratios and trends in crude marriage rates among states between 1967 and 1971 with reduction in crude marriage rates in the states with the relatively high abortion-birth ratios. Change in po licy on induced abortions may be responsible for the increased trend in the U.S. crude marriage rate from 1959-1970 and for its levelling off in 1971 and 1972. Analysis of data also suggested that there is a relation ship between less restrictive abortion policies and a decline in crude marriage rates. An estimation of the number of marriages postponed for at least one year following legal abortion indicated that about 1 abortion in 10 delays a marriage when all the decline in marriage rates were caused by increased abortions and when no legal abortions were subs titutes for illegal abortions. Results were obtained employing general measures for variables, but stronger relationships might be produced if more refined measures which consider race, age, parity, and marital status were introduced.

Selected readings in econometrics from econometrica

The twenty-two articles collected in this book have had and continue to have a considerable impact on the development of econometrics. Since limits must be set in such a broad field, selections have been restricted to articles dealing primarily with statistical methods or applications of statistical methods to economic models.In the study of economics, preference functions vary enormously among individuals; therefore the editors have exercised two choices--balancing articles in econometric methodology with those in econometric application. The result is both discerning and representative, with chapters ranging over such subject matter as the stock market, hybrid corn, international comparison of household expenditure patterns, the demand for durable goods, and the durability of durable goods, as well as the use of econometric models in guiding economic policy and identification problems in model construction. There are also a number of chapters on simultaneous equations: three-state least squares, canonocal correlation theory, alternative estimators, cost of approximate specification, and the statistical implications of a system of simultaneous equations.Both students and professional practitioners of econometrics will find this a convenient reference book and a historical document of the growth and changing emphasis of the subject over the past three decades.Contributors Alfred Cowles, 3rd, R. Frisch and F. V. Waugh, T. Haavelmo, C. F. Roos and V. S. Von Szeliski, H. B. Mann and A. Wald, L. R. Klein, T. C. Koopmans, H. B. Chenery, T. M. Brown, R. L. Basmann, Z. Griliches, H. S. Houthakker, J. Tobin, J. W. Hooper, I. Adelman and F. L. Adelman, R. Stone and D. A. Rowe, T.-C. Liu, F. M. Fisher, A. Zellner and H. Theil, T. J. Rothenberg and C. T. Leenders, M. Nerlove, and G. C. Chow.