多條成長軌跡與非完整資料之研究：SEM與MLM的比較   免費試閱

The Study of Multiple Growth Trajectories and Unbalanced Data: A Comparison of SEM and MLM

在多層次研究如雨後春筍般的蓬勃發展之後，許多學者紛紛研究兩個研究典範：結構方程模式與多層次模式在多層次橫斷面資料分析上的異同，但較少研究探討追蹤資料成長模型在這兩種方法估計結果上的比較。本研究延伸MacCallum等人的研究，將多條成長軌跡延伸到重複觀測時點誤差項變異數不等的情況下，比較結構方程模式與多層次模式在這樣異質性假設以及不完整資料的估計結果。根據437家公司4年度兩組重複觀測資料在這兩種方法上的分析發現，在成長軌跡上得到相同的參數估計值，且在不完整資料的條件下也獲得相似的結果。此結果顯示追蹤資料的多成長軌跡在異質性假設與遺漏資料條件下，這兩種方法可以說是等價。

There are many papers to discuss the similarities and differences between SEM and MLM on the cross-sectional research when the multilevel studies are growing. But there is few paper to investigate their differences in the longitudinal studies. In this paper， we extend the research of MacCallum et al to the heterogeneous hypothesis and compare the results of SEM and MLM under the heterogeneity and unbalance data settings. The findings analyzed from the repeated measures of 437 companies show that the estimates from the latent growth model of SEM and the linear growth model of MLM are the same under heterogeneity and their outcomes of the unbalanced data are similar. In other words， SEM and MLM are equivalent under heterogeneity assumption and missing data in the longitudinal study.