Difference between recursive and non-recursive
We will be dealing mostly with recursive models (no feedback loops
The number of variables in your matrix = the number of paths you are trying
Equation for figuring out the known
SEM Week 2
Path Modeling w/ measured variables
Q: What are the typical statistics that fit an SEM model?
Whats the relationship between covariance and correlation?
-Covariance = two predictors that are correlated with each other more than
Finding () model specification
the correlation between two variables X and Z is equal to the sum of the products
of all paths from each possible valid tracing between X and Z HNDT #1
These tracings, for simple recursive models, include all
Unmeasured variables do not have a scale
-When we set these paths to 1 we tell the program that the ds should have the
same scale as the relevant observed variables (this is known as factor loading)
To adjust output, go to view - Analysis prope
1. Theory. Our model is in-line with school learning theory
2. Time precedence. Causality is unlikely to operate backwards in time. Ability forms
earlier than motivation and achievement
3. Relevant research. Model should be concordant with relevant previo
Single factor latent variable models are tricky. For now, ignore them. In this model, it
would be ethnicity.
It tends to work best when all of your variables are highly correlated with each other.
You need to measure in a concrete way, what is your latent
When comparing nested models, compare the difference between the Chi square tests of
the two models.
-Need to work on finding p-value in models
Reliability is truth over total. V can be gotten by looking at the sample covariance matrix.
Overview and syllabus notes
SEM = multivariate and includes latent variables
-Meant for large surveys
-Better suited for confirmatory analysis (not exploratory)
-Must have specific research questions
-CFA is the measurement of SEM
-First half of class is
Effects of family background (new), ability, motivation and academic coursework (new)
on achievement HNDT #2
In general, there are 6 broad steps for conducting a path analysis
1. Specify the model
2. Check the identification status of the model
SEM Week 4
b = a direct structural effect
c = either a covariance b/w exogenous elements of a variance of an exog element
Reading is exogenous, nothing is predicting it. Its variance comes from perceived
variance of reading