PARALLEL COMPUTING IN STATISTICAL METHODS

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Fen Bilimleri Enstitüsü, İSTATİSTİK ANABİLİM DALI, Türkiye

Tezin Onay Tarihi: 2022

Tezin Dili: İngilizce

Öğrenci: ORÇUN OLTULU

Danışman: Fulya Gökalp Yavuz

Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu

Özet:

Cost-efficient data collection and storage methods enable scientists, companies, and even regular computer users to reach high-dimensional data sets faster and cheaper. Even though personal computers are getting more powerful and efficient, some algorithms, tasks, and problems still require too much computational power and time to run on a personal computer. For a few decades, parallelization in statistical computing had an increasing trend, and researchers put significant effort into converting or adjusting known statistical methods and algorithms in parallel. The main reasons for the transition to parallel methods are the rapid growth in the size and the volume of data and the accelerated hardware developments. In this study, we applied the parallelization technique to statistical algorithms such as Linear Regression models, Non-parametric Regression models, and the measurement error kernel regression operator (MEKRO) algorithm for variable selection in Non-parametric Regression models. Simulation studies are conducted for each algorithm and recorded their accuracy measures and elapsed times to compare and see whether parallelization methods offer significant efficiency while maintaining the accuracy level as high as their sequential versions. The overall simulation results show that parallelization of the offers a great potential of time efficiency with negligible or no changes in accuracy values.