Sample quantiles corresponding to mid p-values for zero–modification tests

Wilson and Einbeck (2015, 2016) propose a test for zero-modification relative to a stated model. The basis of the test is that the number of observed zeros follows a Poisson-binomial distribution. The decision to reject, or otherwise, the non zero–modified model is made by either (i) computing the mid p-value corresponding to the number of observed zeros, or (ii) comparing the number of observed zeros to the relevant “traditional” quantile of the appropriate Poisson–binomial distribution. In general either approach will result in the same decision, but occasionally discrepancies may occur. In this paper we investigate the use of mid-distribution quantiles in approach (ii) above, and show that this reduces the possibility of discrepancies.

Citation

Wilson, P. and Einbeck, J. (2017) Sample quantiles corresponding to mid p-values for zero–modification tests, in Grzegorczyk, M. and Ceoldo, G. (Eds.) Proceedings of the 32nd International Workshop on Statistical Modelling. July 3-7, 2017, Groningen, Netherlands, pp. 275-279.

Type

Conference contribution

Language

en

Collections

Faculty of Science and Engineering