The Annals of Statistics

Directional tests for one-sided alternatives in multivariate models

Abstract

Consider one-sided testing problems for a multivariate exponential family model. Through conditioning or other considerations, the problem oftentimes reduces to testing a null hypothesis that the natural parameter is a zero vector against the alternative that the natural parameter lies in a closed convex cone $\mathscr{C}$. The problems include testing homogeneity of parameters, testing independence in contingency tables, testing stochastic ordering of distributions and many others. A test methodology is developed that directionalizes the usual test procedures such as likelihood ratio, chi square, Fisher, and so on. The methodology can be applied to families of tests where the family is indexed by a size parameter so as to enable nonrandomized testing by $p$-values. For discrete models, a refined family of tests provides a refined grid for better testing by $p$-values. The tests have essential monotonicity properties that are required for admissibility and for desirable power properties. Two examples are given.

Article information

Source
Ann. Statist., Volume 26, Number 6 (1998), 2321-2338.

Dates
First available in Project Euclid: 21 June 2002

https://projecteuclid.org/euclid.aos/1024691473

Digital Object Identifier
doi:10.1214/aos/1024691473

Mathematical Reviews number (MathSciNet)
MR1700234

Zentralblatt MATH identifier
0927.62056

Subjects
Primary: 62H15: Hypothesis testing 62H17: Contingency tables

Citation

Cohen, Arthur; Sackrowitz, Harold B. Directional tests for one-sided alternatives in multivariate models. Ann. Statist. 26 (1998), no. 6, 2321--2338. doi:10.1214/aos/1024691473. https://projecteuclid.org/euclid.aos/1024691473

References

• Agresti, A. and Coull, B. A. (1996). Order restricted tests for stratified comparisons of binomial proportions. Biometrics 52 1103-1111.
• Agresti, A. and Coull, B. A. (1998). Order restricted inference using odds ratios for monotone trend alternatives in contingency tables. Comput. Statist. Data Analy sis. 28, 139-155.
• Agresti, A. and Finlay, B. (1997). Statistical Methods in the Social Sciences, 3rd ed. Prentice Hall, Upper Saddle River, NJ. Baker, J. P., Detsky, A. S., Wesson, D. E., Wolman, S. L., Stewart, S., Whitewell, J., Langer,
• B. and Jeejeebhog, K. N. (1982). Nutritional assessment: a comparison of clinical judgment and objective measurements. New England J. Medicine 306 969-972.
• Berger, V. and Sackrowitz, H. B. (1997). Improving tests of stochastic order. J. Amer. Statist. Assoc. 92 700-705.
• Bhattachary a, B. and Dy kstra (1994). Statistical inference for stochastic ordering. In Stochastic Orders and Their Applications (M. Shaked and J. G. Shantikumar, eds.) 221-249. Academic Press, Boston.
• Cohen, A., Kemperman, J. H. B. and Sackrowitz, H. B. (1994). Projected tests for order restricted alternatives. Ann. Statist. 22 1539-1546.
• Cohen, A., Perlman, M. D. and Sackrowitz, H. B. (1990). Unbiasedness of tests of homogeneity when alternatives are ordered. In Proceedings of the Sy mposium on Dependence in Statistics and Probability (H. W. Block, A. R. Sampson and T. H. Swirts, eds.) 135-146.
• IMS, Hay ward, CA.
• Cohen, A. and Sackrowitz, H. B. (1992). An evaluation of some tests of trend in contingency tables. J. Amer. Statist. Assoc. 87 470-475.
• Cohen, A. and Sackrowitz, H. B. (1994). Association and unbiased tests in statistics. In Stochastic Orders and Their Applications (M. Shaked and J. G. Shantikumar, eds.) 251-274. Academic Press, Boston. Cohen, A. and Sackrowitz, H. B. (1998a). Testing whether treatment is better than control with ordered categorical data: definitions and complete class theorems. Unpublished manuscript. Cohen, A. and Sackrowitz, H. B. (1998b). Testing whether treatment is better than control with ordered categorical data: an evaluation of new methodology. Unpublished manuscript.
• Cohen, A., Sackrowitz, H. B. and Samuel-Cahn, E. (1995). Constructing tests for normal orderrestricted inference. Bernoulli 1 321-333.
• Eaton, M. L. (1970). A complete class theorem for multidimensional one-sided alternatives. Ann. Math. Statist. 41 1884-1888.
• Emerson, J. D. and Moses, L. E. (1985). A note on the Wilcoxon-Mann-Whitney test for 2 × k ordered tables. Biometrics 41 303-309.
• Ferguson, T. S. (1967). Mathematical Statistics, A Decision Theoretic Approach. Academic Press, New York.
• Graubard, B. I. and Korn, E. I. (1987). Choice of column scores for testing independence in ordered 2 × K contingency tables. Biometrics 43 471-476.
• Green, P. J. (1985). Peeling data. In Ency clopedia of Statistical Sciences 6 660-664. Wiley, New York.
• Grove, D. M. (1980). A test of independence against a class of ordered alternatives in a 2 × C contingency table. J. Amer. Statist. Assoc. 75 454-459.
• Kou, S. G. and Ying, Z. (1996). Asy mptotics for a 2 × 2 table with fixed margins. Statist. Sinica 6 809-829.
• Ledwina, T. (1978). Admissible tests for exponential families with finite support. Math. Operationsforsch. Statist. Ser. Statist. 9 105-118.
• Lehmann, E. L. (1986). Testing Statistical Hy potheses. Wiley, New York.
• Marshall, A. W. and Olkin, I. (1970). Inequalities: Theory of Majorization and Its Applications. Academic Press, New York.
• Moses, L. E., Emerson, J. D. and Hosseini, H. (1984). Analy zing data for ordered categories. New England J. Medicine 311 442-448.
• Nguy en, T. T. and Sampson, A. R. (1987). Testing for positive quadrant dependence in ordinal contingency tables. Naval Res. Logist. 34 859-877.
• Pagano, M. and Halvorsen, K. T. (1981). An algorithm for finding the exact significance levels of r by c contingency tables. J. Amer. Statist. Assoc. 76 931-934.
• Patefield, W. M. (1982). Exact tests for trends for ordered contingency tables. J. Roy. Statist. Soc. Ser. C 31 32-43.
• Rahlfs, V. W. and Zimmerman, H. (1993). Scores: Ordinal data with few categories-how they should be analyzed. Drug Inform. J. 27 1227-1240.
• Robertson, T., Wright, F. T. and Dy kstra, R. L. (1988). Order Restricted Statistical Inference. Wiley, New York.
• Wang, Y. (1996). A likelihood ratio test against stochastic ordering in several populations. J. Amer. Statist. Assoc. 91 1676-1683.