Multilevel modelling Flashcards
What is clustering?
students within a class cluster together, employees within an organisation, people within a town, time points
What is a random effect?
parameter varies across groups/clusters, indicating that a teacher or a school or whatever is influencing individual’s performance
What is a fixed effect?
A parameter that is consistent across groups
What is a random intercept model?
where the intercept varies across groups, the slope is still fixed however
What is a random slopes model?
the explanatory effect is different for each group, the r/s between an iv on a dv is diff for different groups,
What is within and between groups variance?
variables have 2 key sources of variance, within and between group variance,
What are the assumptions of multilevel modeling?
Outliers, normality, multicollinearity, linearity, homoscedacity, groups are randomly samples of all possible groups, and also whether mlm is suitable for the rq
How many groups should be used to test in mlm?
Field, kreft de leew, say that at least 20 groups needed
Is it more important to have a lot of individuals within a group or a lot of groups to test?
Can have few people within a group, more important to have a lot of groups
ICC tests test for what?
whether a random intercept is needed, range from 0-1, if larger than .05, need mll, if less, can get away with standard regress, tests whether is sig diff to 0
when to use random slope
spss provides an estimate for variance at each lvl and if sig, retain fixed and random
What are the steps to mlm?
Step 1) test if icc is less than .05 than use standard regress but if not, mlm, step 2) include lvl 1 predictors as fixed effects 3) add random slopes 1 at a time, step 4) include lvl 2 predictors
How do you evaluate individual effects in mlm?
which iv most important for prediction, which iv most important for explaining random component of the model
