1. Provide the value for (using the correct acronym and three decimal places) and
interpret (in no more than one sentence) the comparative fit index, the Tucker-Lewis
index, the root mean square error of approximation and CI90% and p-value for H
C. Standardized factor variance
A way to show that these two models are nested and directly output the Dc2 is to
parameterize the model differently
Recall the r, .84, between Nonverbal Reasoning and Spatial
We may wonder if this r is statistically signif
Model 1 fits just as well as the more complex. For this check out the CMIN number.
E. Rules of thumb for post hoc modifications
1. Examine larger values first
The size of MIs, like c2, are sample dependent
Because we had pretty good fit to start with, an
Pattern Coefficient (pxm)
P = rows
M = columns
If there is not a path between the variable and the factor the direct effect is fixed to zero
Text Output estimates Scalar Regression weights
If given Lambda coefficient matrix, you could
Results may be relative, not absolute
3. Model Generation (MG) exploratory
search for the best fitting, substantively meaningful model, with the same
data, after initial model has been rejected
unacceptable model-data fit r
More common will be to set a loading to one, while at other times you may have to set a
variance to 1.
Common factors to the left, non-common to the right
2. 4 latent constructs: Vc (common variance)
Verbal ability - influences scores on Wo
Latent Means Models
This lecture is all about Latent growth modeling curves.
Means havent been in your model. You can find means in your models with what you
You are interested in latent means, not presumed means.
Here, you want to measure math achi
G = general intelligence
In this model, verbal, non reasoning, spatial, memory are no longer exogenous variables.
You need to add residuals to these variables now. UF1
In the model above u 1 u 8 qould be the eight uniqueness terms.
Confirmatory Factor Analysis
Some of these issues are not unique to CFA, just wanted to get solid footing in path
analysis before introduced to them.
Not really a data reduction technique.
Low correlations may be good in SEM. Depends what you expect to fi
Variance of Tests
1. Test 1: participants read a passage on one page and then are asked to select 1
of 4 images that best illustrates the passage
Test 1: RC + ability to translate something read into a picture
2. Test 2: participants read a passage (e.g.