WK 7 Flashcards
What does a path model allow us to do?
A path model allows us to test several linear models together as a set
What are exogenous variables?
They are essentially independent variables
What are endogenous variables?
They are dependent variables in at least one part of the model
What directions do the arrows go in exogenous variables?
Only have directed arrows going out (basically predictors)
What directions do the arrows go in endogenous variables?
They have directed arrows going in
In a diagram, what does a square represent?
It represents an observed/measured variable
In a diagram, what does a circle represent?
It represents an unobserved/latent variable
In a diagram, what does a two-headed arrow represent?
It represents covariance
In a diagram, what does a single headed arrow represent?
It represents a regression path
In a diagram, when there is a single arrow, what is it showing?
The square that the arrow is pointing in to is the dependent variable/outcome, whereas the other side is the predictor
What does every endogenous variable have?
They have a residual
What do you need after you run a lavaan model?
you need to use a summary function
What is specification in path models?
It concerns which variables relate to which others, and in what ways
What are the standard rules in path models?
- all exogenous variables correlate
- for endogenous variables, we correlate the residuals, not the variables
- endogenous variable residuals do not correlate with exogenous variables
- all paths are recursive (i.e. we cannot have loops like A-> B, B-> A)
What is model identification?
Identification concerns the number of knowns versus unknowns
In order to test our model, what do we need?
We need more knowns than unknowns in order to test our model
What are the knowns?
The knowns are variances and covariances
What are path models based on?
path models are based on the correlation matrix between variables that you have measured
What are the unknowns?
The unknowns are the parameters we want to estimate
What are the degrees of freedom in model identification?
Degrees of freedom are the difference between knowns and unknowns
What are the three levels of identification?
Under-identified, just identified, over-identified models
What are the degrees of freedom for under-identified models?
They have <0 degrees of freedom
What model won’t estimate?
Under-identified models
What are the degrees of freedom for just identified models?
They have 0 degrees of freedom
What is an example of just identified models?
standard linear models
What are the degrees of freedom for over-identified models?
they have > 0 degrees of freedom
What is model estimation?
It refers to finding the ‘best’ values for the unknown parameters
What are the assumption of maximum likelihood estimation?
- large sample size
- multivariate normality
- variables are on a continuous scale
Why does no convergence happen?
- the model is not identified
-the model is mis-specified
-the model is very complex, so more iterations are needed than the program default
What does a statistically significant chi-squared suggest?
It suggests the model does not do a good job of reproducing the observed variance-covariance matrix
What is the range of absolute fit?
ranges from 0 to 1
What is deemed as perfect fit in absolute fit?
0=perfect fit
What is considered good in the range for absolute fit?
values <.05 considered good
What is a perfect parsimony-corrected indices?
0 = perfect fit
What is considered good in the range for parsimony-corrected indices?
values <.05 considered good
What does absolute fit measure?
It measures the discrepancy between the observed correlation matrix and model-implied correlation matrix
What does parsimony-corrected indices do?
Adds a penalty for having more degrees of freedom
What does incremental fit indices do?
It compares the model to a more restricted baseline model
What is the range of a comparative fit infex (CFI)?
Ranges between 0 and 1
What value is the perfect fit, and what values are considered good in comparative fit indices?
1= perfect fit
values > 0.95 considered good
What does the Tucker-Lewis index (TLI) include?
Includes a parsimony penalty
What values are considered good in tucker-lewis index?
Values >0.95 are considered good
What is modification indices?
Modification indices provides the improvement in fit
What do expected parameter changes estimate?
Estimate the value of the parameter were it to be included
How do you extract modification indices and expected parameter chnges?
using summary(model, mod.indices=T)
When is chi-squared statistic significant?
values < 0.05
What is absolute fit?
standardised root mean square residual (SRMR)
What is parsimony-corrected?
root mean square error of approximation (RMSEA)