Abstract

The present study evaluated the performance of robust variance estimation (RVE) and random-effects TSSEM with missing at random (MAR) data under realistic conditions. The performance of the two methods was compared in terms of the first stage of MASEM entailing the pooling of correlation estimates and in the second stage when the SEM model is estimated. Findings from this study indicated that RVE and TSSEM provided unbiased pooled correlations at the first stage of MASEM. And TSSEM and RVE both provided unbiased parameter estimates at the second stage. With MAR missing correlations, RVE provided acceptable coverage rates only in scenarios with at least 100 studies while TSSEM provided acceptable coverage rates even with a small to a moderate number of studies (k = 25 and 50).

摘要

本研究评估了稳健方差估计 (RVE) 和随机效应 TSSEM 在现实条件下随机缺失 (MAR) 数据的性能。两种方法的性能在 MASEM 的第一阶段进行了比较，该阶段需要合并相关性估计，第二阶段是估计 SEM 模型。这项研究的结果表明，RVE 和 TSSEM 在 MASEM 的第一阶段提供了无偏的汇总相关性。并且 TSSEM 和 RVE 在第二阶段都提供了无偏参数估计。由于 MAR 缺少相关性，RVE 仅在至少有 100 项研究的情况下提供可接受的覆盖率，而 TSSEM 提供可接受的覆盖率，即使是少量到中等数量的研究（k  = 25 和 50）。