在這篇文章中，我們提供一些用以生成 Dirichlet 隨機向量的演算法，並根據以下標準來評估這些演算法的表現：（一）電腦生成時間；（二）敏感度；以及（三）適合度。另外，我們特別檢驗一個基於 beta 變量轉換的演算法，並提供三個方針以減少此演算法的生成時間。模擬的結果顯示，除了所有（或大部分）的形狀變數都相當接近零的情況之外，基於我們所提出的方針整合而成的演算法顯著地在電腦生成時間上勝過其他的演算法。

In this article, we describe various well-known Dirichlet generation algorithms and evaluate their performance in terms of the following criteria: (i) computer generation time, (ii) sensitivity, and (iii) goodness-of-fit. In addition, we examine in particular an algorithm based on transformation of beta variates and provide three useful guidelines so as to reduce its computer generation time. Simulation results show that the proposed algorithm outperforms significantly other approaches in terms of computer generation time, except in cases when all (or most) shape parameters are close to zero.

[1] Ahrens, J. H. and Dieter, U. (1974). “Computer Methods for Sampling from Gamma, Beta, Poisson and Binomial Distributions.” Computing, 12, 223-246.
[2] Ahrens, J. H. and Dieter, U. (1982). “Generating Gamma Variates by a Modified Rejection Technique.” Communications of the ACM, 25, 47-54.
[3] Alam, K., Abernathy, R., and L., W. C. (1993). “Multivariate Goodness-of-Fit Tests Based on Statistically Equivalent Blocks.” Communications in Statistics. Theory and Methods, 22, 1515-1533.
[4] Anderson, T. W. (1966). “Some Nonparametric Procedures Based on Statistically Equivalent Blocks.” Proceedings of International Symposium on Multivariate Analysis, pp. 5-27.
[5] Best, D. J. (1983). “A Note on Gamma Variate Generators with Shape Parameter Less Than Unity.” Computing, 30, 185-188.
[6] Chatfield, C. and Goodhardt, G. (1975). “Results Concerning Brand Choice.” Journal of Marketing Reseach, 12, 110-113.
[7] Cheng, R. (1977). “The Generation of Gamma Variables with Non-integral Shape Parameters.” Applied Statistics, 26, 71-75.
[8] Connor, R. J. and Mosimann, J. E. (1969). “Concepts of Independence of Proportions with a Generalization of the Dirichlet Distribution.” Journal of the American Statistical Association, 64, 194-206.
[9] Fishman, G. (1978). Principles of Discrete Event Simulation. John Wiley & Sons Inc.
[10] Foutz, R. V. (1980). “A Test for Goodness-of-Fit Based on Empirical Probability Measure.” Annals of Statistics, 8, 989-1001.
[11] Fraser, R. V. (1957).NonparametricMethods in Statistics.Wiley Series in Statistics.
[12] Gupta, R. D. and Richards, D. S. P. (2001). “The History of the Dirichlet and Liouville Distributions.” International Statistical Review, 69, 433-446.
[13] Hogg, R. V. and Craig, V. A. (1978). Introduction to Mathematical Statistics (4th Edition). Macmillan & Company, New York.
[14] Huillet, T. and Paroissin, C. (2009). “Sampling from Dirichlet Partitions: Estimating the Number of Species.” Environmetrics, Doi: 10.1002/env.977.
[15] Hung, Y. C., Balakrishnan, N., and Lin, Y. T. (2009). “Evaluation of Beta Generation Algorithms.” Communications in Statistics - Simulation and Computation, 38, 750-770.
[16] IMSL (1980). InternationalMathematical and Statistical Libraries. Houston, Texas.
[17] J ¨ohnk, M. D. (1964). “Erzeugung von Betaverteilten und Gammaverteilten Zufallszahlen.” Metrika, 8, 5-15.
[18] Kotz, S., Balakrishnan, N., and Johnson, N. L. (2000). Continuous Multivariate Distributions, Volume 1, Methods and Applications (2nd Edition). Wiley''s Series in Probability and Statistics.
[19] Lange, K. (2005). “Applications of Dirichlet Distribution to Forensic Match Probabilities.” Genetica, 96, 107-117.
[20] Loukas, S. (1984). “Simple Methods for Computer Generation of Bivariate Beta Random Variables.” Journal of Statistical Computation and Simulation, 20, 145-152.
[21] Madsen, R. E., Kauchak, D., and Elkan, C. (2005). “ModelingWord Burstiness Using the Dirichlet Distribution.” Proceedings of the 22th International Conference on Machine Learning, pp. 545-552.
[22] Marsaglia, G. and Tsang,W.W. (2001). “A SimpleMethod for Generating Gamma Variables.” ACM Transactions on Mathematical Software, 26, 363-372.
[23] Martin, J. J. (1967). Bayesian Decision Problem and Markov Chains. John Wiley & Sons, New York.
[24] Mosimann, J. E. (1962). “On the Compound Multinomial Distribution, the Multivariate Beta-Distribution and Correlations among Proportions.” Biometrica, 49, 65-82.
[25] Narayanan, A. (1990). “Computer Generation of Dirichlet Random Vectors.” Journal of Statistical Computation and Simulation, 36, 19-30.
[26] Sakasegawa, H. (1983). “Stratified Rejection and Squeeze Method for Generating Beta Random Numbers.” Annals of the Institute of Statistical Mathematics Part B, 35, 291-302.
[27] Saltelli, A., Tarantola, S., Compolongo, F., and Ratto, M. (2004). Sensitivity Analysis in Practice: A Guide to Accessing Scientific Models. John Wiley & Sons.
[28] Schmeiser, B. W. and Babu, A. J. G. (1980). “Beta Variate Generation via Exponential Majorizing Functions.” Operations Research, 28, 917-926.
[29] Serfling, R. J. (1980).Approximation Theorems forMathematical Statistics.Wiley''s Series in Probability and Statistics.
[30] Tanizaki, H. (2008). “A Simple Gamma Random Number Generator for Arbitrary Shape Parameters.” Economic Bulletin, 3, 1-10.
[31] Tsui, K. W., Matsumura, E. M., and Tsui, K. L. (1985). “Multinomial-Dirichlet Bounds for Dollar Unit Sampling in Auditing.” Accounting Review, 60, 76-96.
[32] Tukey, J. W. (1947). “Non-Parametric Estimation II. Statistically Equivalent Blocks and Tolerance Regions - The Continuous Case.” The Annals of Mathematical Statistics, 18, 529-539.
[33] Wilks, S. S. (1962).Mathematical Statistics. Princeton University Press.
[34] Zechner, H. and Stadlober, E. (1993). “Generating Beta Variates via Patchwork Rejection.” Computing, 50, 1-18.