Part1: multilevel models Flashcards
What is Longitudinal data?
Longitudinal data is repeated observations of the same variables collected from the same individual(s) over multiple time points (over a period of time). It’s useful for analyzing changes and trends over time within the same individual(s). It accounts for individual changes, thus minimizing the variability that comes with following changing individuals.
what is Cross-sectional data?
Collecting one measurement per subject to basically get a picture of the group in current time, meaning we see no time changes etc. Can be described as a picture of group at a given time point.
What is a mixed effects model?
A model that contains an expression for both fixed and random effects.
What is level 1 of the mixed effects model?
Changes within individuals over time
What is level 2 of the mixed effects model?
Changes between individuals (different at beginning of the study? Do they evolve in the same ways?)
What are the distributional assumptions of mixed effects models?
Normal distribution of the random effects
What is the difference between a marginal model and a Hierarchical/nested model?
Hierarchical: Data can be nested within higher-level units. The models estimates the effects of predictors at each level and allows for random effects.
Marginal:
The marginal model is the average of all individuals meaning that the marginal model is always within the hierarchical model!! You cannot have a hierarchical model without a marginal but you can have a marginal without a hierarchical model!
What is BLUP?
Best linear unbiased prediction. Used to predict random effects in mixed effects models
What does a homoscedastic model mean?
A statistical model in which the variance of the errors (residuals) is constant across all levels of the independent variables.
Why bootstrap fixed effects estimates instead of using the ones computed by the lmer?
- Bootstrap has no normality assumption
- Good at capturing variability in data
- Often both are used and evaluated
Calculate standard deviations for residuals
Random effects residual * parameter estimates (low and high)
What does a heteroscedastic mixed-effects mean?
It is a model that allows different variances across different groups/data.
If we want to test if the variance of the residuals in the high and low dose groups are the same?
REML-based likelihood ratio tests.
How do calculate inverse probability weighting?
1/P(Response) where P(Response) = Responses/Number of possible responses
