Multilevel Linear Models Flashcards
Hierarchical Data
Variables are often nested within other variables, which leads to correlations between data from the same context. For example, school children are nested in classrooms.
Intraclass Correlation
Data from the same context will be more similar than data from different contexts. This violates the assumption of independence therefore creating issues in ANOVA and Regression.
Benefits of MLM
Removes some of the assumptions of other statistical methods (eg independence), robust to violations of normality, can cope with missing data points, includes correlations between individuals as part of the analysis.
Fixed Coefficients
The intercept and slope cannot change and remain the same across all contexts. In standard analyses, coefficients are fixed.
Random Coefficients
The intercept and slope are free to vary across contexts.
Assessing MLM Fit
AIC, AICC, BIC and ACIC are four statistics that can be used. Otherwise, examine the change in -2LL as a way of comparing models using a chi square distribution.
Covariance Structure
Specifies the form of the variance-covariance matrix. In a standard analysis, this cannot change. In MLM, this needs to be selected. Default is usually ok as long as the slope isn’t random.
MLM in SPSS
Information Criteria - model fit statistics
Tests (and estimates) of Fixed Effects - whether the fixed factors and intercept are significant
Covariance Parameters - does the random intercept vary significantly across contexts