Scholarlink

<p>Analysis of high-dimensional survey data is difficult because of multicollinearity, redundant predictors, and the need to account for complex sampling. This study aimed to adapt survey-weighted variable screening, specifically sure independence screening (SIS), for use with complex survey designs and evaluate its performance in identifying childhood diarrhea predictors. Data were obtained from the multiple indicator cluster survey (MICS), which included a sample of 8923 children. The analysis focused on diarrhea using MICS data as a case study. SIS was implemented with survey weights, redundancy checks, iterative refinement, and replicate-weight stability and compared with principal component filtering (PCF) using the AUC Brier Score, and AIC. Compared with PCF, SIS achieved superior discrimination, calibration, and fit while maintaining interpretable covariates. These findings demonstrate that survey-weighted SIS provides a computationally efficient and interpretable method for dimensionality reduction in complex surveys, outperforming component-based alternatives.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Statistical methods for dimensionality reduction in complex surveys: application of survey-weighted sure independence screening

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月17日

查看全文
  • Ali Satty

摘要

Analysis of high-dimensional survey data is difficult because of multicollinearity, redundant predictors, and the need to account for complex sampling. This study aimed to adapt survey-weighted variable screening, specifically sure independence screening (SIS), for use with complex survey designs and evaluate its performance in identifying childhood diarrhea predictors. Data were obtained from the multiple indicator cluster survey (MICS), which included a sample of 8923 children. The analysis focused on diarrhea using MICS data as a case study. SIS was implemented with survey weights, redundancy checks, iterative refinement, and replicate-weight stability and compared with principal component filtering (PCF) using the AUC Brier Score, and AIC. Compared with PCF, SIS achieved superior discrimination, calibration, and fit while maintaining interpretable covariates. These findings demonstrate that survey-weighted SIS provides a computationally efficient and interpretable method for dimensionality reduction in complex surveys, outperforming component-based alternatives.