SEM Flashcards
What is the goal of CFA?
(Brown and Moore, 2014)
To establish the number and nature of latent variables that account for the variation and covariation among a set of observed measures
After conducting CFA why are the relations between factors theoretically free of measurement error?
A major advantage to this type of analysis is that when relations among factors are examined, the relations are theoretically free of measurement error because the error has been estimated and removed, leaving only common variances (Ullman, 2006).
What is a latent variable?
A latent variable is an unobservable variable that influences more than one DV and which accounts for the correlations among these DVs.
- i.e. a common cause of multiple DVs
Number of factors should be ___than the number of variables?
less
What does CFA require the researcher to do a priori?
CFA requires researchers to explicitly specify all characteristics of the hypothesised measurement model to be examined.
The researcher should have a strong conceptual or empirical foundation to guide the specification and evaluation of models.
Alternative to CFA?
EFA - a more data driven approach (Gallagher and Brown, 2013)
What does it mean when a model is identified?
Identification refers to the concept that a CFA solution can be estimated only if the number of freely estimated parameters does not exceed the number of data points in the input matrix.
Model evaluation
The model you have specified is computed for different parameter estimates (variances and covariances).
The best fitting parameters are determined - These are defined as the values that produce a model covariance matrix that is closest to the empirical covariance matrix
What is a method of parameter optimisation during model evaluation?
Maximum likelihood attempts to find the model parameter estimates that maximize the probability of observing the available data if the data were collected from the same population again. (Likelihood function)
How to compute the goodness of fit?
Chi-squared
Several fit indices should be used as they all provide different information about the model fit (Brown, 2006)
What is model modification?
Searching for the presence or absence of localised strain in the solution (i.e. specific points of ill-fit)
Wald Test?
Wald Test determines if any parameters can be removed without significantly reducing the quality of fit
Lagrange Multiplier?
Asks if the model fit would be improved significantly by adding one more parameter.
Last stage in building an SEM?
The interpretability, size and statistical significance of the model’s parameter estimates should be examined. For example, the parameters should be of a magnitude and direction that is in accord with conceptual or empirical reasoning.