Karteikarten
Multi Level Modelling
Karten in diesem Stapel (11)
- 1What is Multi level Modelling?A flexible framework to model relationships between variables in situations where data are clustered ( Classified groups in the same sample, i.e participant form cluster 1).
– Multilevel regression,
– Multilevel path analysis,
– Multilevel CFA, etc. - 2What is clustering?• Sometimes referred to as nesting or hierarchies, data may be referred to as having hierarchical or nested structure
* Clustering is common:
– Students within classrooms ( School, classroom, student)
– Employees within an organisation (Admin, Team leader, manager) - 3When and why would you use multilevel modelling?To know the effects of the cluster group within a sample. E.g. So if you look at cluster one, (state school, low socio economic status and single parent family). This cluster tells us something different to Cluster 2. We can compare at the cluster level.
- 4What are the advantages of using MLM?It allows us to test random effects (the differences of each cluster)
- 5What happens if we have evidence of a random effect?We encounter Cross level interactions. E.g. IV: Teacher experience, level 1 unexperienced, level 2 highly experience. Each level interacts with DV thus causing cross level interactions.
- 6What is random effects in MLMThe effects of each cluster - differences of individual clusters not usually tests in standard MR
- 7What are fixed effects in MLM?One line of best fit to describe the correlation in MR
- 8What test do we use to get a non negative R2 value?Snijder and Boskers (2012) formula produces a non negative R2 value.
- 9What are the problems with effect size in MLM?* complex
* Can result in unexplained variance increasing when II's are added.
* Between and within group effects are confounded unless predictors are centred. - 10What are intra class correlations?ICC is the ratio of variance between groups to variance within groups. This is used to evaluate the suitability of MLM approaches for your dataset. If the ICC equals 0 use standard multiple regression (This means there is no random effects of the clusters so you should just use MR for Fixed effect) If the ICC is greater than .05 then use MLM
- 11What are the assumptions of MLM?*All assumptions of MR apply - outliers, normality, multicollinearity, linearity and homoscadecity.
* Correlated predictors are problematic in MLM. At .5 and .6 Medium correlation is problematic.
* Clusters must be random samples.
*Random effects are normally distributed.
* Suitability of MLM for dataset (ICC test)
*Sample size
– For power, increasing number of groups at Level 2 more important than increasing individuals within groups
Informationen
- Beschreibung
- Karteikarten am Multi Level Modelling, erstellt von seonapalmer am 30/09/2015.
- Erstellt von
- seonapalmer
- Aufrufe
- 4
- Bewertung