Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
30 Cards in this Set
- Front
- Back
What is the purpose of CFA?
|
To test hypotheses about the factor structure that underlies a set of observed variables
|
|
What is the difference between EFA and CFA?
|
EFA is data-driven, and factor structure is interpreted post hoc from the results
CFA is theory-driven, where factor structure is specified a priori, the program only calculates specified loadings (all else = 0) |
|
When should you use CFA?
|
When you want to test a priori hypotheses based on theory and/or previous research with these variables
|
|
What is the range for factor loadings?
|
0 to +/- 1
|
|
In a CFA diagram, what shape represents latent factors?
|
Circles/ Ovals
|
|
What is a latent factor?
|
Unmeasured constructs that are made up of common variance among items
|
|
What two bits of information do you need to provide to the software package to run a CFA?
|
1. Variables to be included in the analysis
2. Specified model: factor loadings and correlations |
|
What is a FIXED parameter?
|
A relationship between factors which is constrained to be zero.
|
|
What does CFA compare in order to find fit of the model?
|
Estimated Variance-Covariance Matrix
and Actual Variance-Covariance Matrix |
|
What are the 4 steps in CFA?
|
1. prepare for analysis
2. evaluate model fit 3. evaluate factor loadings and factor correlations 4. evaluate alternative models |
|
What scale should data be measured on in CFA?
|
Interval scale
|
|
Why do you need to have at least one constrained parameter?
|
If no parameters are constrained, the observed and estimated variance-covariance matrices will be equal, and you will be unable to assess the fit of your hypothesised model
|
|
What is the ideal number of participants in a CFA or EFA?
|
More than 10 participants per estimated parameter
|
|
What does a good model fit mean?
|
The hypothesised model provides an accurate account of observed relationships in the dataset
|
|
Which type of parameter can never contribute to bad fit?
|
Estimated Parameters
|
|
Why does good fit NOT mean that the model is correct?
|
Because it does not rule out alternatives
|
|
Why does good fit NOT mean that the factors explain a lot of variance?
|
Because a model that explains very little variance can still fit well if the researcher has correctly identified constraints
|
|
What does the chi-squared test do?
|
Tests the degree of similarity between estimated and actual variance-covariance matrices
|
|
What does a non-significant chi-squared test indicate?
|
That the matrices are not significantly different, meaning that the model fits well
|
|
What are the limitations of the chi-squared test?
|
1. It is sensitive to sample size
2. Assumption of multivariate normality is often violated |
|
What are the two options you have if the chi-squared test is significant?
|
1. calculate the chi-squared/df ratio
2. if other indices suggest good fit, downplay chi-squared test |
|
What do absolute and incremental fit indices show?
|
How much of the observed variance-covariance matrix has been accurately accounted for by the model
|
|
What do residual fit indices show?
|
The discrepancies between the estimated and observed variance-covariance matrices
|
|
What is the threshold for good fit in absolute and incremental fit indices?
|
>.95 (the higher the better)
|
|
What is the threshold for good fit in residual fit indices?
|
SRMR: <.80
RMSEA: <.60 (the lower the better) |
|
How many absolute and incremental fit indices should be reported?
|
Two
|
|
What does a bad fit mean?
|
the model does not account for all relationships in data. One or more fixed parameters needs to be freed.
|
|
When comparing the fit of two models, what should you do?
|
Nest the models. The model with fewer parameters is nested within the model with more parameters
|
|
What test should you perform when comparing the fit of nested models?
|
The chi-squared difference text
|
|
If the Chi-squared difference test is significant, what does this mean?
|
The nested model fits worse than the larger model, thus the larger model is preferred because it provides a better account of the observed relationships
|