Journals / Bilecik Şeyh Edebali Üniversitesi Fen Bilimleri Dergisi / 2022 / Cilt: 9 - Sayı: 1

Comparison of Penalized Regression Methods through a Simulation Study

Bir Simülasyon Çalışması ile Cezalı Regresyon Yöntemlerinin Karşılaştırılması

Pages
80–91
Publication Date
2022-06-30
DOI
—
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

Simulation studies show that the structure of the real coefficient vector has a significant effect on the model performance revealed by the methods.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Penalized regression methods are often used to obtain stable coefficient estimates in case of multicollinearity problems in the dataset.
📊
Key Findings & Data:In this study, a detailed comparison of the performances of ridge, LASSO, elastic net and adaptive LASSO penalized regression methods, which are widely used in the literature, is made through simulation studies depending on the structure of the real coefficient vector.
💡
Conclusion & Impact:Simulation studies show that the structure of the real coefficient vector has a significant effect on the model performance revealed by the methods.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:Doğrusal Regresyon · Ridge · Lasso

Abstract

Penalized regression methods are often used to obtain stable coefficient estimates in case of multicollinearity problems in the dataset. In addition, these methods can make automatic variable selection depending on the nature of the penalty term applied. In this study, a detailed comparison of the performances of ridge, LASSO, elastic net and adaptive LASSO penalized regression methods, which are widely used in the literature, is made through simulation studies depending on the structure of the real coefficient vector. Mean squared error on the test set, misclassification rate, false positive rate and active set sizes are used as comparison criteria in the study. Simulation studies show that the structure of the real coefficient vector has a significant effect on the model performance revealed by the methods.

Özet

Veri kümesinde çoklu iç ilişki problemi olması durumunda kararlı katsayı tahminleri elde etmek için sıklıkla cezalı regresyon yöntemleri kullanılır. Ayrıca bu yöntemler uygulanan ceza teriminin yapısına bağlı olarak otomatik değişken seçimi de yapabilmektedir. Bu çalışmada literatürde yaygın kullanım alanı bulan ridge, LASSO, elastik net ve uyarlanabilir LASSO cezalı regresyon yöntemlerinin gerçek katsayı vektörünün yapısına bağlı olarak simülasyon çalışmaları yoluyla performanslarının ayrıntılı olarak karşılaştırılması yapılmıştır. Çalışmada karşılaştırma kriteri olarak test kümesi üzerinde hata kareler ortalaması, yanlış sınıflama oranı, yanlış pozitif oranı ve aktif küme büyüklükleri kullanılmıştır. Simülasyon çalışmaları, gerçek katsayı vektörünün yapısının yöntemlerin ortaya çıkardığı model performansı üzerinde önemli etkisinin olduğunu göstermektedir.

🔖

Cite This Article

GENÇ, M. (2022). Comparison of Penalized Regression Methods through a Simulation Study. Bilecik Şeyh Edebali Üniversitesi Fen Bilimleri Dergisi, Cilt: 9 - Sayı: 1, 80–91. https://ajindex.com/articles/475485

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (16)

  1. Montgomery, D. C., Peck, E. A. & Vining, G. G. (2021). Introduction to linear regression analysis, John Wiley & Sons. Hoerl, A. E. & Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics12 (1), 55-67.
  2. Rao, C. R. & Toutenburg, H. (1995).
  3. Linear models, Springer. Sarkar, N. (1992). A new estimator combining the ridge regression and the restricted least squares methods of estimation. Communications in statistics-theory and methods21 (7), 1987-2000.
  4. Kaçıranlar, S., Sakallıoğlu, S., Akdeniz, F., Styan, G. P. & Werner, H. J. (1999). A new biased estimator in linear regression and a detailed analysis of the widely-analysed dataset on Portland cement. Sankhyā: The Indian Journal of Statistics, Series B, 443-459.
  5. Özkale, M. R. & Kaçıranlar, S. (2007). The restricted and unrestricted two-parameter estimators. Communications in Statistics-Theory and Methods36 (15), 2707-2725.
  6. Miller, A. (2002).
  7. Subset selection in regression, CRC Press.
  8. Frank, L. E. & Friedman, J. H. (1993). A statistical view of some chemometrics regression tools. Technometrics35 (2), 109-135.
  9. Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological)58 (1), 267-288.
  10. Zou, H. & Hastie, T. (2005). Regularization and variable selection via the elastic net. Journal of the royal statistical society: series B (statistical methodology)67 (2), 301-320.
  11. Zou, H. (2006). The adaptive lasso and its oracle properties. Journal of the American statistical association 101(476), 1418-1429.
  12. Sirimongkolkasem, T., & Drikvandi, R. (2019). On regularisation methods for analysis of high dimensional data. Annals of Data Science 6(4), 737-763.
  13. Meinshausen, N., & Yu, B. (2009). Lasso-type recovery of sparse representations for high-dimensional data. The annals of statistics 37(1), 246-270.
  14. Yüzbaşı, B., Arashi, M., & Ejaz Ahmed, S. (2020). Shrinkage Estimation Strategies in Generalised Ridge Regression Models: Low/High Dimension Regime. International Statistical Review 88(1), 229-251.
  15. Ahmed, S. E., Kim, H., Yıldırım, G., & Yüzbaşı, B. (2016). High-Dimensional Regression Under Correlated Design: An Extensive Simulation Study. In International Workshop on Matrices and Statistics (pp. 145-175).
  16. Springer, Cham. Shahriari, S., Faria, S., & Gonçalves, A. M. (2015). Variable selection methods in high-dimensional regression—A simulation study. Communications in Statistics-Simulation and Computation 44(10),