Regularized population proportion estimation under ranked set sampling: a ridge logistic regression approach
Ridge regression is an effective approach for addressing multicollinearity in statistical estimation by stabilizing regression coefficients when multiple auxiliary variables are highly correlated. This article proposes a hybrid ridge logistic regression-assisted model for finite population proportion estimation under concomitant-based Ranked Set Sampling (CRSS). The proposed model combines observed responses with predictions from a ridge logistic regression, thereby exploiting multiple auxiliary variables while mitigating the effects of multicollinearity. A bootstrap resampling scheme designed for Ranked Set Sampling is adopted for variance estimation and statistical inference. The performance of the proposed model is evaluated through extensive Monte Carlo simulations under varying sample sizes, correlation structures, and dimensions of auxiliary variables, including additional moderate levels of multicollinearity. Applications to the Body Fat and Wisconsin Diagnostic Breast Cancer datasets are presented to demonstrate its practical utility. The results generally show lower bias and mean squared error and higher relative efficiency than existing estimators, particularly under severe multicollinearity. The findings demonstrate the effectiveness of integrating ridge-regularized logistic regression with CRSS for finite population proportion estimation using multiple auxiliary variables.
Authors
- Alok Kumar (ORCID: https://orcid.org/0000-0002-5554-6612)
- Raghaw Raman Sinha (ORCID: https://orcid.org/0000-0001-6386-1973)
- H. P. Singh
Institutions
- Dr. B. R. Ambedkar National Institute of Technology Jalandhar (IN)
- Vikram University (IN)
Publication Details
- Journal
- Scientific Reports
- Published
- 2026-11-20
- DOI
- https://doi.org/10.1038/s41598-026-80178-w
- Primary Topic
- Survey Sampling and Estimation Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00