Small Sample Estimation of Structural Equation Model Based on SAM

本文提出一种基于结构后测量(SAM)的结构方程模型(SEM)小样本估计方法，旨在解决传统结构方程模型在小样本情况下的估计偏差问题。SAM方法通过分阶段建模策略，先估计测量模型，再估计结构模型，结合普通最小二乘法(OLS)的预测优势与SAM的稳健性，有效提升小样本下的参数估计精度。通过蒙特卡洛模拟实验以及教育投入对人才质量培养的仿真实验验证，在小样本情况下，所提出的方法在路径系数和载荷矩阵的估计上都具有更高的准确性，研究结果为小样本社会科学研究提供了方法论支持。This paper proposes a small sample estimation method of structural equation model (SEM) based on structural after measurement (SAM), aiming to solve the estimation bias of traditional SEM in small samples. Through the phased modeling strategy, the SAM method first estimates the measurement model and then the structural model, which combines the prediction advantage of ordinary least squares (OLS) and the robustness of SAM to effectively improve the accuracy of parameter estimation under small samples. Through the Monte Carlo simulation experiment and the simulation experiment of education investment on talent quality training, the proposed method has higher accuracy in the estimation of path coefficient and load matrix in small samples, and the research results provide methodological support for social science research with small samples.