본 연구는 다층 모형을 활용하는 연구자들이 흔히 2수준 다층모형의 사용에 편중되어 있다는 점에 근간하여, 수집된 자료의 구조적 특성에 따라 어떠한 다층모형을 활용하는 것이 보다 적절한 분석 방법이 될 것인지 파악하는 것을 목적으로 하였다. 이를 위해 2수준 다층모형과 3수준 다층모형, 이원 다층모형의 모수 추정방식, 제공되는 정보 유형 및 유사점과 차이점을 이론적 차원에서 제시하였으며, 경험적 자료 분석 결과를 통해 세 모형의 모수 추정치의 차이를 비교·논의하였다.

[닫기]

The purpose of this study was to provide the theoretical relationship of parameter estimation between 2-level HLM, 3-level HLM, and crossed multilevel model(CMM) and to show empirical illustrations of the estimates when the three models analyzed the same data. The three models were compared under three conditions of model specification. The results of the analyses were as below. First, the three models decomposed variance components differently when the base models were compared. CMM decomposed the total variance into four components, while 2-level HLM and 3-level HLM decomposed the variance into only two and three components respectively. Second, the three models yielded nearly equal parameter estimates of the fixed effects when the predictors are specified at macro-level only. Third, 2-level HLM showed larger parameter estimates for fixed effects, and more shrunken standard errors of the estimates than CMM when the predictors were specified at both micro- and macro-level in the models. However, the results of parameter estimation of 3-HLM and CMM were almost identical.

[닫기]