6: Factor Analysis And Latent Variables Flashcards
What are patent variables?
Constructs we can’t measure directly, so must infer.
What is the inferred common cause of several “indicator” variables?
A latent variable.
What is the factor analysis term for Beta values?
Factor loadings.
In latent path models, what are “other” causes?
Unique aspects of indicators, not related to the latent variable, or error.
What is rotational indeterminacy?
When the model is under-identified, there are infinite solutions and we cannot solve the factor loadings.
Why is the breadth and meaning of indicators so crucial?
The looser the relationships between indicators, the broader the interpretation of the latent variable.
Describe the method of triads for estimating correlations by hand.
Pick 2 correlations involving X, then divide by the correlation needed to cancel out the other paths.
Where do the axes come from in a factor matrix?
They come from the “numbers” before any rotation.
What should each factor be like in “simple structure”?
Should have some high loadings and many small loadings, so defined by a clear and distinguishable set of variables.
What should each variable be like in “simple structure”?
Should ideally load mainly on one factor and have minimal loadings on other factors.
What type of analysis is used to test a path model explicitly?
Confirmatory factor analysis.
What type of analysis is used to let data suggest number and meaning of factors?
Exploratory factor analysis.