Quantitative 3 Flashcards
(14 cards)
What is the effect size?
A factor loading, which is correlation (only after some corrections have been applied) between on observed variable and a factor
Range of matrices came out of a favour analysis but typically we interpret what is called the pattern matrix
What is payers matrix??
Associated with what is called on oblique rotation
What is obliques
Used when you expect you expect your factors to intercorrelate
How high should factor listings be??
- no universal agreement regarding acceptance effective sizes
- rule of thumb
- rough estimate, factor loading interest should exceed 0.30
Structural Equation Modelling (SEM)
Stimulatianeously estimated multiple dependence relationship ( to multiple, multiple regression)
Look at multiple IVs + DVs
Take predictive as well as measure models into account
Confirmatory Factor Analysis
CFA’s comprise the buildings blocks of SEM
CFA is also a data analytic technique in its own right
Example CFA
Two different theoretical model:
- single-factor model:
One theory suggest that the KFT measures. (General intelect)
- three-factor model: Another theory suggests that the KFT measures three separates dimensions of mental ability: Verbal Non- verbal Quantitative abilities
Single Factor (CFA model)
One latent trait
Six observed variables
Each observed variables had an associated error term to take measurement error into account
Arrow point down from the latent variables to each observed variables because the idea is that latent variables causes responses on the observed variable
Goodness - of - fit - indices
Comparative for index (CFI) - should be greater 0.95
Tucker Lewis index (TLI) - should be greater 0.95
Root mean square error of approximation (RMSEA) should be less 0.05
Heywood cases:
In admissible parameter estimate i.e stuff happening in your model that doesn’t make sense
Common expanse includes:
- negative residual variance
- correlations > | 1 |
What to do if you have Heywood cases?
Change/check your model because it might be misspecified
Get a larger sample size
Try adding more indicated per latent variable
Relationship Between latents
Exogenous variables - latent variables that has one or more arrows emanating from it on IV
Endogenous variable - latent that had one or more arrows pointing at it (DV)
Latent Variables can potentially act as both IVs and DVs
SEM
Possibilities with SEM are endless
Versatile way to test you hypotheses and models
SEM brings together factor analysis, multiple regression and another technique called path analysis and rolls them all into one integrated system
Factor analysis:
Aka: Common factor analysis Standard factor analysis Exploratory factor analysis Principal components analysis
Takes observed variables (e.g items) scales and groups them together empirically
