Evaluating the Impact of Resampling Methodologies on Multiple Regression Performance in Housing Price Analysis: A Comprehensive Empirical Study of Bootstrap, Stratified, and Systematic Sampling Techniques

Elsegai, Dr.Heba Mahmoud; El Daba, Dr.Amr

doi:10.21608/rijcs.2025.402309.1367

Evaluating the Impact of Resampling Methodologies on Multiple Regression Performance in Housing Price Analysis: A Comprehensive Empirical Study of Bootstrap, Stratified, and Systematic Sampling Techniques

نوع المستند : المقالة الأصلية

المؤلفون

¹ كلية التجارة جامعة المنصورة

² Department of Economics The Arab Academy for Banking & Financial Sciences

10.21608/rijcs.2025.402309.1367

المستخلص

This study compares bootstrap, stratified, and systematic sampling for housing price regression models. Using both simulated and King County real-world data, we evaluate their impact on predictive accuracy through adjusted R².
Key Findings
Stratified sampling outperformed other methods consistently, showing 8.3% higher R² than bootstrap and 12.7% higher than systematic sampling. This advantage was strongest for smaller samples (n<1000).
Method Comparison
Bootstrap sampling showed small-sample bias risks, while systematic sampling displayed high variability with clustered data. Stratified sampling maintained superior performance across all sample sizes.
Real-World Validation
King County data confirmed stratified sampling's advantages were even greater in practice than simulations, highlighting its real-world applicability.
Implications
Results strongly recommend stratified sampling for real estate modeling, particularly with limited data. Findings emphasize sample representativity's crucial role in regression accuracy, with broader applications to naturally stratified variables. The study contributes to the growing body of literature on sampling methodology optimization and provides empirical evidence for best practices in regression model development.

الكلمات الرئيسية