传统的问卷的结构效度和测量等价性研究中，由于数据中交叉载荷的存在，常常无法获得良好的模型拟合。本文主要聚焦于数据中存在交叉载荷的测量模型，通过Monte Carlo模拟研究比较了验证性因子分析和探索性结构方程模型在处理存在交叉载荷的数据时二者在结构效度分析和进行因子载荷水平的测量等价性检验时的异同，并考察了二者实际应用于NEO-FFI量表时的表现。结果表明：（1）当不存在交叉载荷时，CFA的模型拟合和参数估计都要略优于ESEM，但二者没有显著差异。存在交叉载荷时，ESEM在CFI、TLI和RMSEA等模型拟合指标上的表现都要显著优于CFA，并且受到交叉载荷的大小和数量影响较小。对于模型的主要参数的估计，ESEM对因子相关的估计的精度要显著高于CFA，而且能够估计交叉载荷，但其估计精度有待提高；而CFA对于因子载荷的估计精度要高于ESEM，但二者差异不明显。此外，当样本量较小时，ESEM会遇到相对严重的收敛问题，可以考虑优先使用CFA。（2）在正确识别存在组间差异的因子载荷的能力上CFA要优于ESEM。观测变量数量越大，CFA和ESEM的识别能力越低。较大的样本量、载荷差异数量和大小都有助于提高二者的识别能力。卡方差异检验的正确识别比率最高，CFI差异检验次之。ESEM和CFA都存在错误识别组间差异的问题，ESEM错误识别比率略高于CFA，但二者差异不明显。从检验方法的角度来看，在CFA上，卡方差异检验的错误识别比率远远高于CFI差异检验和RMSEA差异检验；而对ESEM来说，CFI的错误识别比率则最低。（3）通过基于ESEM的测量等价性检验，NEO-FFI量表在性别、年龄和教育水平三个维度上都获得了测量等价性的支持；在东方文化背景下，NEO-FFI是一个可以广泛使用的量表，研究者可以基于观察数据进行跨性别、年龄和教育水平的平均数比较。

In traditional studies about construct validity and measurement invariance, researchers always could get satisfying model fitting indices because of the existence of cross loading. This study mainly compares the performances of CFA and ESEM on construct validity and measurement invariance when dealing with the data with cross loadings through Monte Carlo simulation study. At the same time, performances of them on the application to questionnaire NEO-FFI are also been evaluated. The results indicated that,(1)Results revealed that CFA had better fit indices and estimate accuracy than ESEM when applied to data sets without cross loadings, but there is no significant difference between them. On the contrary, when with cross loadings, ESEM had better fit indices than CFA, including CFI, TLI and RMSEA, and ESEM was almost unaffected by the size and ratio of cross loadings in the model. Also, ESEM had a significant advantage of CFA in estimate accuracy on parameters of factor correlation and cross loading. However, it performed worse on factor loadings, although the results were very close. Besides, CFA was suggested when the sample size is too small, because ESEM would face a worse convergence problem. (2)CFA had a better performance on identifying the different factor loadings between groups. With bigger number of observed variables, CFA and ESEM performed worse, however, the performance of both of them improved when the sample size, the number of different observed variables and the difference between groups increased.ESEM and CFA both have the problems of misidentification between groups, and ESEM performed worse than CFA. However, the difference is not significant. From the perspective of test method, Chi square difference test has the highest ratio of misidentification on CFA, and CFI has the lowest on ESEM. (3)The measurement invariances of NEO-FFI on CUs, factor loadings, partial intercepts, uniquenesses, factor variance-covariance over gender, age and education were supported through ESEM measurement invariance test. Under the eastern cultural background, NEO-FFI could be a widely used questionnaire and researchers are able to compare the means of the factors over gender, age and education on the basis of observed data.