EEG-Based Sleep Stage Classification — Meta-Analysis Summary & Reference Implementation

This dataset accompanies the doctoral thesis "A Comparative Meta-Analysis of Feature Extraction Methods Used in EEG-Based Sleep Stage Classification" (Demirci, Selçuk University Institute of Natural and Applied Sciences, 2026). It contains the per-stage F1 extraction CSVs (single-reviewer strict and four-source consensus), the publication-ready summary workbook with 16 sheets, the reference Python implementation that reproduces the pooled F1 values reported in Table 4.7, the pseudocode for Appendix EK-11, and a README describing the file structure and reproducibility steps. The pooled estimates use a DerSimonian-Laird random-effects model on the logit scale with a 1/sqrt(N) standard error proxy (Peters et al., 2006) and IntHout 95% prediction intervals. All five sleep stages (W, N1, N2, N3, REM) are reproduced within 0.05 absolute tolerance of the thesis values.本数据集配套于博士学位论文《基于脑电图（Electroencephalogram, EEG）睡眠分期分类的特征提取方法比较元分析》（Demirci，塞尔丘克大学自然与应用科学学院，2026年）。本数据集包含逐睡眠分期提取的F1分数CSV文件（含单审稿人严格审核标准结果与四来源共识结果）、包含16个工作表的可直接用于发表的汇总工作簿、可复现论文表4.7中报告的合并F1分数的参考Python实现、附录EK-11所用的伪代码，以及一份说明文件结构与复现步骤的README文档。本次合并估计采用DerSimonian-Laird随机效应模型，以logit尺度结合1/√(N)标准误代理（Peters等，2006年），并使用IntHout 95%预测区间。全部五个睡眠分期分别为觉醒（Wake, W）、N1、N2、N3及快速眼动睡眠（Rapid Eye Movement, REM），其复现结果与论文数值的绝对误差均不超过0.05。