1
Q
latent classes
A
categories of theta
within each class, items are independent (local independence)
2
Q
goal
A
classify subjects to the latent classes on the basis of observed item scores
3
Q
conditional probabilities
A
single points, since theta is categorical
4
Q
maximum likelihood
A
- expectation maximization algorithm
E-step:
> determine expected values of the latent classes given some initial values for the other parameters
M-step:
> maximize the likelihood using these expected values to obtain new values for the other parameters
- iterate between these steps - newton-raphson algorithm
> approximates the log-likelihood function locally, using linear functions to find firections towards the maximum
5
Q
identification
A
- scaling the LV
> the LV has a scale that is defined by the number of categories - statistical identification
> k should not exceed M
> this can happen if you have too few observed variables in your model
> k = #cond.probs + #class.probs-1
6
Q
model fit
A
absolute fit measures
1. expected number of subjects
> overall.prob * N
2. goodness of fit
> pearson X^2, G^2
comparative fit measures
> likelihood ratio, AIC, BIC
7
Q
X2
A
should decrease as model complexity increases
