Data And Code For The Paper Titled 'Variance-Calibrated Cross-Individual Bootstrapping for Small-Sample Neuroscience'

This repository contains the MATLAB source code, simulation results, and empirical datasets associated with the manuscript of the same name, submitted to Scientific Reports.

Abstract

Small sample sizes (N≤10) are a pervasive limitation in experimental neuroscience. This project introduces CIB-VC, a novel bootstrapping framework designed for hierarchical datasets (few subjects, many trials). By combining cross-individual trial recombination with a variance calibration step, CIB-VC achieves nominal coverage and higher statistical power than traditional hierarchical methods.

Repository Structure

The repository relies on a flat file structure for the main analysis code and processed data, with a specific subfolder for the raw biological recordings.

1. Main Directory (Root)

Analysis & Result Generation Scripts

generate_simulation_data.m: The core Monte Carlo simulation engine. It generates the comparative performance metrics for CIB-VC vs. Stratified and Hierarchical bootstrapping.

Note: Running this script takes significant computational time.

robustness.m: Generates Robustness/Dot Plot.

efficiency.m: Generates Efficiency Map.

powerVStypeI.m: Generates Power vs. Type-I Error Plot.

Biological Data:

fte.mat: Processed Frequency Tracking Error (FTE) metrics derived from the raw recordings.

all_conditions.mat: Metadata encoding the experimental conditions corresponding to the FTE data.

Pre-computed Simulation Results Outputs of generate_simulation_data.m, provided to allow immediate figure reproduction:

CIB_VC_sim_summary.csv

CIB_VC_power_summary.csv

CIB_VC_type1_summary.csv

2. Raw Data Folder (/raw_fish_data)

This folder contains the original, individual data files for the N=5 Eigenmannia virescens subjects before preprocessing into fte.mat. These are provided for transparency and archival purposes.

amasra.mat

ardahan.mat

erzincan.mat

gaziantep.mat

samsun.mat

Reproducibility GuideSystem Requirements

MATLAB: R2021b or later.

Toolboxes: Statistics and Machine Learning Toolbox.

Warning: This process performs extensive Monte Carlo iterations (R=1500) across multiple scenarios and distributions. It will overwrite the existing CSV files upon completion.

Licensing & AttributionIf you use this method or code in your research, please cite the associated Scientific Reports manuscript:

Uyanik, I. (2026). Variance-Calibrated Cross-Individual Bootstrapping for Small-Sample Neuroscience, Scientific Reports.

ContactIsmail Uyanik Hacettepe University Department of Electrical and Electronics Engineering Ankara, Türkiye

本仓库包含提交给《Scientific Reports》的同名稿件相关的MATLAB源代码、仿真结果与经验数据集。

摘要
小样本量（N≤10）是实验神经科学领域普遍存在的局限。本项目提出CIB-VC，一种针对层级式数据集（少量被试、大量试次）的新型自举（bootstrapping）框架。通过跨个体试次重组与方差校准步骤的结合，CIB-VC能够达到标称覆盖性，且相较于传统层级方法具备更高的统计效力。

仓库结构
本仓库针对主分析代码与预处理后数据采用扁平化文件结构，并为原始生物记录数据设置专属子文件夹。

1. 主目录（根目录）
分析与结果生成脚本
generate_simulation_data.m：核心蒙特卡洛（Monte Carlo）仿真引擎，用于生成CIB-VC与分层、层级自举方法的对比性能指标。
注意：运行该脚本需耗费大量计算时间。
robustness.m：生成稳健性分析散点图。
efficiency.m：生成效能热力图。
powerVStypeI.m：生成统计效力与I型错误（Type-I Error）关系图。

生物数据集：
fte.mat：源自原始记录的预处理频率追踪误差（Frequency Tracking Error, FTE）指标。
all_conditions.mat：对应FTE数据的实验条件元数据。

generate_simulation_data.m的预计算仿真结果输出文件，可直接用于复刻实验图表：
CIB_VC_sim_summary.csv
CIB_VC_power_summary.csv
CIB_VC_type1_summary.csv

2. 原始数据文件夹（/raw_fish_data）
本文件夹包含N=5条绿色裸背电鳗（Eigenmannia virescens）被试的原始个体数据文件，该数据尚未被预处理为fte.mat。本文件的公开旨在保障研究透明度与归档需求。
amasra.mat
ardahan.mat
erzincan.mat
gaziantep.mat
samsun.mat

可复现性指南
系统要求
MATLAB：R2021b及更高版本。
工具箱：统计与机器学习工具箱。

警告：本流程将在多场景与分布下执行大量蒙特卡洛迭代（R=1500），运行完成后将覆盖现有CSV文件。

授权与归属
若您在研究中使用本方法或代码，请引用相关《Scientific Reports》稿件：
Uyanik, I. (2026). Variance-Calibrated Cross-Individual Bootstrapping for Small-Sample Neuroscience, Scientific Reports.

联系方式
伊斯梅尔·乌亚尼克 哈杰泰普大学电气与电子工程系 土耳其安卡拉