On Analysis of Two-Way Contingency Tables with some Sampling Zero Cell Counts — Application to Human Reproduction Wastage according to Age —

A systematic procedure is described for the analysis of I x J cross-classifications with some nonstructural zeros. This includes (1) use of pseudo-Bayes estimators to remove zeros (2) fitting the smoothed array to all possible configurations of the log-linear models (3) fitting the data to the quasi -independence model after omitting counts of some anomalous cells — on the basis of standardized residuals to detect the cause of deviation from the null independence model. As a guide, a risk function value is computed for each array of smoothed pseudo-Bayesian estimates.

Real-life data (based on a sampling survey conducted on the population of female nurses in Kuwait) is used for illustration.

Es wird ein systematisches Verfahren zur Analyse von I × J Kreuzklassifikationen mit einigen nieht-strukturellen Nullen beschrieben.

Dies schließt ein 1. die Benutzung von Pseudo-Bayes-Schätzern zur Beseitigung der Nullen; 2. Die Anpassung der geglätteten Tafel an alle möglichen Konfigurationen der log-linearen Modelle; 3. Die Anpassung der Daten an das Quasi-Unabhängigkeitsmodel nach Weglassen anomaler Zellen — auf der Grundlage standardisierter Residuen —, um die Ursache der Abweichung von der Null-Hypothese der Unabhängigkeit zu entdecken. Als Richtlinie wird der Wert einer Risikofunktion für jede Tafel mit geglätteten Pseudo-Bayes-Schätzern berechnet.

Der Analysengang wird an Daten, die bei allen Krankenschwestern in Kuwait erhoben wurden, erläutert.

Key-Words:

Two-Way Contingency Tables - Pseudo-Bayes Estimators - Log-Linear Models - Sampling Zero Cell Counts - Quasi-Independence Model

Schlüssel-Wörter:

Zweiweg-Kontingenztafeln - Pseudo-Bayes-Schätzer - Log-lineare Modelle - mit Null besetzte Zellen - Quasi-Unabhängigkeitsmodell

• References

• 1 Bishop Y. M. M, Fienberg S. E, Holland P. W. Discrete Multivariate Analysis: Theory and Practice. M. I. T. Press, Cambridge: Mass; 1975
• 2 Demtng W. E, Stephan P. F. On a least squares adjustment of a sampled frequency table when the expected marginal totals are known. Ann. Math-Statist 1940; 11: 427-444.
• 3 Good I. J. A Bayesian significance test for multinomial distributions (with discussion). J. Roy. Statist. Soc. Ser. B 1967; 29: 399-431.
• 4 Goodman L. A. The analysis of cross-classified data: independence, quasi-independenee, and interactions in contingency tables with or without missing values. J. Amer. Statist. Assoc 1968; 63: 1091-1131.
• 5 Goodman L. A. The multivariate analysis of qualitative data: interactions among multiple classifications. J. Amer. Stat. Assoc 1970; 65: 226-256.
• 6 Goodman L. A. The analysis of multidimensional contingency tables, stepwise procedure and direct estimation methods for building models for multiple classifications. Technometrics 1971; 13: 33-61.
• 7 Haberman S. J. The analysis of residuals in cross-classified tables. Biometrics 1973; 29: 205-220.
• 8 Ibrahim M-E, Hathout H. M, Moussa M. A. A, A/Razik M. On factors contributing to reproductive wastage — An analytical study on the nursing staff in Kuwait. 1980. (In press).
• 9 Sutherland M. Estimation in large sparse multinomials. Ph. D. dissertation, Department of Statistics, Harvard University. 1974
• 10 Sutherland M, Holland P, Fienberg S. E. Combining Bayes and frequency approaches to estimate a multinominal parameter. In Studies in Bayesian Econometrics and Statistics, edited by Fienbero S. E, Zellner A. Amsterdam; North Holland: 1974
• 11 Upton G.J. G. The Analysis of Cross-tabulated Data. John Wiley & Sons, Chichester; New York: 1978