Robust Bayesian Estimation of Mixed Normal Dirichlet Models to Study the Effect of Some Climatic Factors on Evaporation

Ayoob, Hassan

doi:10.25673/122140

Proceedings of International Conference on Applied Innovation in IT · 2025/08/29 · Vol. 13 · Issue 4 · pp. 367–374

Robust Bayesian Estimation of Mixed Normal Dirichlet Models to Study the Effect of Some Climatic Factors on Evaporation

Hassan Sami and Asmaa Ayoob

📄 Download PDF DOI: 10.25673/122140

Abstract

This study proposes and validates a robust Bayesian model based on a Dirichlet process mixture of normals (DMNM) for probability density estimation and missing data imputation in multivariate datasets. The primary focus is on addressing the challenge of incomplete data by providing a flexible and accurate estimation of their underlying probability density function. To fit the model, three Bayesian estimation algorithms are implemented and compared: the Expectation-Maximization (EM) algorithm, the Markov Chain Monte Carlo (MCMC) method, and a Traditional Bayesian (TB) algorithm. The framework is applied to real-world climatic data (temperature, humidity, wind speed, and evaporation) obtained from the Meteorological Service in Basra, Iraq, with artificially introduced missing values at rates of 10%, 20%, and 40%. Model performance is evaluated using two key metrics: the Mean Squared Error (MSE) for imputation accuracy and computational execution time. The results demonstrate that the EM algorithm achieves the highest estimation accuracy (lowest MSE), while the TB method is the most computationally efficient. This work provides a practical toolkit for the statistical analysis of incomplete multivariate data in fields such as environmental modeling, hydrology, and agriculture.

Keywords

Bayesian Estimation Dirichlet Process Mixture Normal Model Gibbs Sampler Climate Factors.

References

T. Peltola, “Dirichlet Process Log Skew-Normal Mixture with a Missing-at-Random Covariate,” Econometrics, vol. 11, no. 4, p. 24, 2023, doi: 10.3390/econometrics11040024.
T. S. Ferguson, “A Bayesian analysis of some nonparametric problems,” Ann. Stat., vol. 1, no. 2, pp. 209–230, 1973.
S. Heydari, Bayesian Nonparametric Dirichlet Process Mixture Modeling in Transportation Safety Studies, Ph.D. dissertation, Univ. of Waterloo, Waterloo, ON, Canada, 2017.
P. Cardoso, J. M. Dennis, J. Bowden, B. M. Shields, T. J. McKinley, and MASTERMIND Consortium, “Dirichlet process mixture models to impute missing predictor data in counterfactual prediction models: An application to predict optimal type 2 diabetes therapy,” Univ. of Exeter, Exeter, U.K., Res. Rep., 2024.
T. P. Minka, Estimating a Dirichlet distribution, Tech. Rep., Massachusetts Inst. of Technology, Cambridge, MA, USA, 2000.
J. Lin, On the Dirichlet distribution, M.S. thesis, Queen’s Univ., Kingston, ON, Canada, 2016.
D. Heinz, Hyper Markov non-parametric processes for mixture modeling and model selection, Ph.D. dissertation, Carnegie Mellon Univ., Pittsburgh, PA, USA, 2010.
C. Archambeau, Dirichlet processes, Lecture Notes, Univ. College London, London, U.K., 2008.
Z. Li, L. Wu, W. Zhou, and S. Ghosh, “Multivariate density estimation with missing data,” Dept. of Statistics, North Carolina State Univ., Raleigh, NC, USA, Res. Rep., 2018.
A. Gelman, J. Gabry, D. Simpson, A. Vehtari, and M. Betancourt, “Bayesian workflow,” Bayesian Anal., vol. 16, no. 1, pp. 1–51, Jan. 2021.
D. J. Spiegelhalter, K. Abrams, and J. P. Myles, “Bayesian methods in health technology assessment: A review,” Lancet, vol. 397, no. 10280, pp. 243–251, 2021.
S. Richardson et al., “Bayesian statistics and its modern applications,” Philos. Trans. R. Soc. A, vol. 379, no. 2196, Art. no. 20200146, Jan. 2021.
C. Robert, “Being Bayesian in the 2020s: Opportunities and challenges,” Philos. Trans. R. Soc. A, vol. 379, no. 2196, Art. no. 20200244, 2021.
T. L. Griffiths, K. A. Hockema, and J. B. Tenenbaum, “Bayesian models of cognition: Applications in cognitive science,” Trends Cogn. Sci., vol. 25, no. 3, pp. 192–205, Mar. 2022.
M. van de Schoot et al., “Bayesian statistics and modelling for psychological research: A tutorial,” Psychol. Methods, vol. 27, no. 2, pp. 200–221, 2022.
T. J. Loredo, “Bayesian astrostatistics: A backward look to the future,” Statist. Sci., vol. 35, no. 3, pp. 437–458, 2020.
G. Papaspiliopoulos and O. Papaspiliopoulos, “Grand challenges in Bayesian computation,” arXiv:2410.00496, 2024.
A. Kruschke and R. Liddell, “The Bayesian new statistics: Hypothesis testing, estimation, meta-analysis, and power analysis from a Bayesian perspective,” Psychon. Bull. Rev., vol. 29, pp. 3–30, 2022.
R. McElreath, Statistical Rethinking: A Bayesian Course with Examples in R and Stan, 2nd ed. Boca Raton, FL, USA: CRC Press, 2020.
D. Hussein, Estimation of the fuzzy reliability function using the Bayesian method with an application, M.S. thesis, Al-Basrah Univ., Basrah, Iraq, 2019, unpublished.
H. Abdulhadi, Estimating the risk function of the new composite distribution (Exponential–Akash) using Bayesian estimation methods, M.S. thesis, Al-Basrah Univ., Basrah, Iraq, 2023, unpublished.
H. Salem, Comparison of the maximum likelihood method and the genetic method with the Bayesian methods for estimating the survival function of an expanded power function distribution with application, M.S. thesis, Al-Basrah Univ., Basrah, Iraq, 2022, unpublished.
A. A. Yaqoub, Time-varying coefficients estimation for the dynamic model in case of discrete survival data with application, Ph.D. dissertation, Univ. of Baghdad, Baghdad, Iraq, 2021, unpublished.
A. Hussein, Estimation of the skew-normal distribution model using the expectation–maximization algorithm, M.S. thesis, Al-Mustansiriyah Univ., Baghdad, Iraq, 2020, unpublished.