Performance of Confidence Intervals for Population Mean of Poisson Distribution with Data from the Renewal Process

Main Article Content

Patchanok Srisuradetchai Chotika Rasikun

Abstract


Count data are commonly found in a wide range of disciplines from medical to social sciences. Regarding an interval estimation for a mean, researchers tend to assume that count data have a Poisson distribution, in which the mean and variance are equal and the interarrival time has an Exponential distribution. A plausible explanation is that the confidence interval for the Poisson mean is simpler to compute and readily found in virtually all statistical programs. However, when count data are generated from a renewal process, the interarrival time of count data can possess any distribution. This paper, therefore, aims to study the performance of seven confidence intervals constructed for the Poisson mean, provided interarrival time is set to be Gamma or Weibull distribution. Because of the given distributions, variance of count data shows to be either greater or smaller than the average. Count data generated from a negative Binomial model is also investigated. The findings show that the Index of Dispersion (ID) can be used to predict the performance of the confidence intervals. If ID is less than or equal to one, all confidence intervals produce the coverage probabilities greater than or equal to the desired confidence level (0.95). On the contrary, if ID is greater than one, the resulting coverage probabilities are lower than the desired confidence level. Moreover, an increase in the sample size is found to have a marginal effect on the coverage probability.


Keywords

Article Details

Section
Applied Science Research Articles

References

[1] G. Boshnakov, T. Kharrat, and I. G. McHale, “A bivariate Weibull count model for forecasting association football scores,” International Journal of Forecasting, vol. 33, pp. 458–466, 2017.

[2] R. P. Dobrow, Introduction to Stochastic Processes with R. Hoboken, New Jersey: John Wiley & Sons, 2016.

[3] R. Baker and T. Kharrat, “Event count distributions from renewal processes: Fast computation of probabilities,” IMA Journal of Management Mathematics, vol. 29, no. 4, pp. 415–433, 2018.

[4] T. Kharrat, G. Boshnakov, I. G. McHale, and R. Baker, “Flexible regression models for count data based on renewal processes: The countr package,” Journal of Statistical Software, 2018.

[5] B. McShane, M. Adrian, E. T. Bradlow, and P. S. Fader, “Count models based on weibull interarrivals times,” Journal of Business & Economic Statistics, vol. 26, no. 3, pp. 369–378, 2008.

[6] R. Winkelmann, “Duration dependence and dispersion in count-data models,” Journal of Business & Economic Statistics, vol. 13, no. 4, pp. 467–474, 1995.

[7] E. L. Lehmann and P. Romano, Testing Statistical Hypotheses. New York: John Wiley & Sons, 2005.

[8] L. A. Barker, “A comparison of nine confidence intervals for a Poisson parameter when the expected number of events is ≤ 5,” The American Statistician, vol. 56, no. 2, pp. 85–89, 2002.

[9] M. Tanusit, “Two-Side confidence intervals for the Poisson means,” International Journal of Modeling and Optimization, vol. 2, no.2, pp. 589–591, 2012.

[10] M. Khamkong, “Approximate confidence interval for the mean of Poisson distribution,” Open Journal of Statistics, vol. 2 no. 2, pp. 204–207, 2012 (in Thai).

[11] Y. Guan, “Moved score confidence intervals for means of discrete distributions,” American Open Journal Statistics, vol. 1, pp. 81–86, 2001.

[12] G. Upton and I. Cook, Oxford Dictionary of Statistics, 2nd ed. Oxford University Press, 2006.

[13] R. Sultan and S. P. Ahmad, “Posterior estimates of Poisson distribution using R software,” Journal of Modern Applied Statistical Methods, vol. 11, no. 2, pp. 530–535, 2012.

[14] C. A. Rohde, Introductory Statistical Inference with the Likelihood Function, 1st ed. London: Springer, 2014.