Why You Should Always Include a Random Slope for the Lower-Level Variable Involved in a Cross-Level Interaction
read more
Citations
Social Isolation and Psychological Distress During the COVID-19 Pandemic: A Cross-National Analysis.
Beyond t test and ANOVA: applications of mixed-effects models for more rigorous statistical analysis in neuroscience research
Beyond t test and ANOVA: applications of mixed-effects models for more rigorous statistical analysis in neuroscience research
Political trust and the relationship between climate change beliefs and support for fossil fuel taxes : Evidence from a survey of 23 European countries
Multilevel Models for the Analysis of Comparative Survey Data: Common Problems and Some Solutions
References
R: A language and environment for statistical computing.
Hierarchical Linear Models: Applications and Data Analysis Methods
Hierarchical Linear Models: Applications and Data Analysis Methods.
Multilevel analysis : an introduction to basic and advanced multilevel modeling
Data Analysis Using Regression and Multilevel/Hierarchical Models
Related Papers (5)
Centering predictor variables in cross-sectional multilevel models: a new look at an old issue.
Understanding Interaction Models: Improving Empirical Analyses
Frequently Asked Questions (9)
Q2. What is the reason why the authors prefer to measure accuracy in terms of the coverage rate?
The reason why the authors prefer to measure accuracy in terms of the coverage rate is that the standard error is a (downward) biased estimator of the sampling distribution standard deviation in small samples.
Q3. What should researchers do to avoid omissions of random slope terms?
12Going beyond the case of cross-national surveys, researchers using mixed-effects models to analyse other types of multilevel data should similarly make sure that their conclusions about cross-level interactions and lower-level effects do not hinge on the omission of the corresponding random slope terms.
Q4. How many linear mixed-effects models are there?
For each of the 30 cross-level interactions (5 dependent variables 6 lower-level predictors), the authors estimate two specifications, resulting in a total of 60 linear mixed-effects models.
Q5. What is the main effect of omitting the random slope term?
Omitting the random slope term associated with a cross-level interaction will not, in general, introduce systematic bias into coefficient estimates,3 but it will lead to overly optimistic statistical inference for the cross-level interaction term and the coefficient (i.e., the ‘main effect’) of the lower-level variable involved in the interaction.
Q6. What is the significance of the omission of the random slope term?
Even if the omission of the random slope term does not lead to a change in statistical significance, it will lead to standard errors that are too small and confidence intervals that are too narrow.
Q7. What is the proportion of confidence intervals that do not cover the true effect size?
The proportion of 95 per cent confidence intervals that do not cover the true effect size (i.e., the actual coverage rate) is generally smaller than the nominal rate, and often by a substantial margin.
Q8. What is the significance of the residuals for lower-level observations belonging to the same cluster?
the residuals for lower-level observations belonging to the same cluster are highly positively correlated when they have similar values on xij and zjxij.
Q9. What is the main effect of the lower-level component of the cross-level interaction term?
This model is widespread in applied research,but the above analysis shows that it is misspecified and provides anticonservative inference for the cross-level interaction term and the main effect of its lower-level component.