How can high-tech manufacturing achieve high innovation productivity? A configurational path analysis under the TOE framework.

1. What is this dataset? This repository contains the comprehensive dataset and original execution scripts (in R and Python) supporting the dynamic Qualitative Comparative Analysis (QCA) of high-tech manufacturing innovation productivity in China. It provides all necessary materials to fully reproduce the configurational path analysis, temporal trend visualizations, industry heterogeneity evaluations, and out-of-sample predictive validity tests presented in the manuscript based on the Technology-Organization-Environment (TOE) framework. 2. How was this dataset collected? The raw panel data were collected from Chinese A-share listed high-tech manufacturing firms covering the period from 2015 to 2024. Financial and patent data were sourced from authoritative databases including CSMAR and WIND. 3. What files are included? The repository is structured into 6 core files to ensure complete transparency and reproducibility: PANELDATA.csv: The primary panel dataset containing the foundational data for the analytical sample, used as the main input for the dynamic QCA process. DYNAMIC.R: The core R script utilizing the QCA and admisc packages. It executes the fuzzy-set calibration, necessity and sufficiency analyses (truth table minimization), and computes both between-group and within-group consistencies across different industry configurations. Calibrated_Data.csv: The fully calibrated fuzzy-set dataset exported from the main QCA procedure, serving as the direct input for the out-of-sample testing. Out-of-Sample Predictive Validity Test.py: A Python script utilizing pandas and seaborn to perform predictive validity testing on a holdout sample (2020-2024). It calculates the consistency and coverage of the specific configurations and automatically generates scatter plots for validation. plot_data.csv: A highly structured dataset specifically extracted and formatted from the QCA clustering results, dedicated to generating temporal trend lines. photo.R: An R script utilizing the ggplot2 package to read plot_data.csv and visualize the intertemporal evolutionary trends of configurational consistency over the decade. 4. How can this dataset be used? Researchers and reviewers can download this complete package into a single local directory to achieve "plug-and-play" reproducibility. By running the R and Python scripts sequentially, users can replicate the exact configurational pathways, robustness checks, and high-quality figures discussed in the study. Furthermore, it serves as a methodological template for scholars intending to integrate dynamic QCA with machine-learning-inspired out-of-sample prediction in management research.1. 本数据集为何物？ 本仓库包含支撑中国高科技制造业创新生产率动态定性比较分析(Qualitative Comparative Analysis, QCA)的完整数据集与原生执行脚本（基于R与Python语言）。其提供了所有必要材料，可复现基于技术-组织-环境框架(Technology-Organization-Environment, TOE)的手稿中所述的组态路径分析、时序趋势可视化、行业异质性评估与样本外预测有效性检验全流程。 2. 本数据集如何采集？ 原始面板数据采集自2015至2024年期间的中国A股上市高科技制造企业。财务与专利数据源自国泰安（CSMAR）、万得（WIND）等权威数据库。 3. 本仓库包含哪些文件？ 为确保全流程透明性与可复现性，本仓库共包含6个核心文件： PANELDATA.csv：核心面板数据集，承载分析样本的基础数据，为动态定性比较分析的主要输入数据源。 DYNAMIC.R：核心R脚本，依托QCA与admisc包实现，可执行模糊集校准、必要性与充分性分析（真值表极小化），并计算不同行业组态的组间与组内一致性。 Calibrated_Data.csv：经主QCA流程导出的完整校准模糊集数据集，作为样本外测试的直接输入数据源。 Out-of-Sample Predictive Validity Test.py：依托pandas与seaborn库的Python脚本，用于对2020-2024年的留存样本执行预测有效性测试，可计算特定组态的一致性与覆盖度，并自动生成验证用散点图。 plot_data.csv：从QCA聚类结果中专门提取并格式化的高结构化数据集，专用于生成时序趋势线。 photo.R：依托ggplot2包的R脚本，用于读取plot_data.csv并可视化十年来组态一致性的跨时期演化趋势。 4. 本数据集可如何使用？ 研究人员与评审人员可将完整套件下载至本地单一目录，实现“即插即用”式的可复现研究。通过依次运行R与Python脚本，用户可复现研究中提及的精确组态路径、稳健性检验与高质量图表。此外，本套件可作为方法学模板，供有志于在管理研究中整合动态QCA与机器学习范式样本外预测的学者使用。