Please enable JavaScript.

Coggle requires JavaScript to display documents.

CFA (Issues for CFA (Identification (If number of variables =n (if no. of…

- - - - large sample sizes mean that even small effects are significant
      - we emphasise testing the whole model rather than individual effects
      - p-value could change if we use a different method to estimate the model parameters
      - greater general awareness of issues wit significance testing
  - - - might need to transform data
  - - - no. of observation = n(n+1)/2
      - if no. of free parameters is equal to the no. of obs = Just-identified
        
        df=0
      - If there are more free parameters than obs = underidentified
        
        df<0
      - if less free parameters than obs = overidentified
        
        df>1
    - - correlating the factors with another factor
  - - - assumes multivariate normality
    - - unweighted least squares
      - Generalised least squares
  - - - discrepancy between model covariance matrix and sample covariance matrix can be reasonably attributed to sampling error?
      - chi-square
    - - Absolute fit statistics
        
        RMR
        
        SRMR
      - comparative
        
        CFI
      - Parsimony-adujsted
        
        RMSEA
- - - - rectangle
    - - ellipse
    - - circle
    - - double headed arrow
    - - single headed arrow