02b Structural Equation Models with Directly Observed Variables
Structural Equation Models with
Directly Observed Variables II
James G. Anderson, Ph.D.
Purdue University
Identification
• Over Identified
• Just Identified
• Under Identified
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 19
DF = 2
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 20
DF = 1
gpa
academic
1
error1
height
weight
attract
1
error2
rating
Over Ideridentified Model
Number of variances/covariances = 21
No. of parameters estimated =20
DF =1
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Just Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 21
DF = 0
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Just Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 21
DF = 0
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Under Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 22
DF = - 1
Covariances
•
•
•
•
Among observed variables
Among exogenous variables
Among measurement errors
Among errors in the equations
Covariances among Observed
Variables
rowtype
varname
n
perform
knowl
98
value
98
sat
98
train
98
98
cov
perform
0.022
cov
knowl
0.017
0.053
cov
value
0.024
0.028
0.121
cov
sat train
0.004
0.004
-0.006
0.09
cov
train
0.018
0.018
0.035
-0.006
0.094
0.058
1.379
2.877
2.461
2.117
mean
Covariances among Exogenous
Variables
gpa
academic
1
error1
height
weight
attract
1
error2
rating
Underidentified Model
Number of variances/covariances = 21
No. of parameters estimated =22
DF =0
Covariances among Measurement
Errors
var_a
var_a
var_p
eps1
eps2
1
1
anomia67
1
eps3
powles67
path_p
67
alienation
1
anomia71
powles71
1
path_p
71
alienation
ses
1
educatio
eps4
1
1
zeta1
var_p
SEI
1
1
delta1
delta2
Example 6: Model C
Exploratory analysis
W heaton (1977)
Model Specification
1
zeta2
Covariances among the Errors in
the Equations
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 20
DF = 1
Class Exercise
• Create a new model:
– From the menu choose File/New
• Specify the Data file:
– Choose File/Data Files
– Browse to the tutorial folder. The path is:
• C:\Program Files\Amos 6\Examples
• In the Files of type list select SPSS
• Select Fels_mal.sav
• Estimate the Parameters of the different models
and compare their fit statistics
Directly Observed Variables II
James G. Anderson, Ph.D.
Purdue University
Identification
• Over Identified
• Just Identified
• Under Identified
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 19
DF = 2
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 20
DF = 1
gpa
academic
1
error1
height
weight
attract
1
error2
rating
Over Ideridentified Model
Number of variances/covariances = 21
No. of parameters estimated =20
DF =1
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Just Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 21
DF = 0
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Just Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 21
DF = 0
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Under Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 22
DF = - 1
Covariances
•
•
•
•
Among observed variables
Among exogenous variables
Among measurement errors
Among errors in the equations
Covariances among Observed
Variables
rowtype
varname
n
perform
knowl
98
value
98
sat
98
train
98
98
cov
perform
0.022
cov
knowl
0.017
0.053
cov
value
0.024
0.028
0.121
cov
sat train
0.004
0.004
-0.006
0.09
cov
train
0.018
0.018
0.035
-0.006
0.094
0.058
1.379
2.877
2.461
2.117
mean
Covariances among Exogenous
Variables
gpa
academic
1
error1
height
weight
attract
1
error2
rating
Underidentified Model
Number of variances/covariances = 21
No. of parameters estimated =22
DF =0
Covariances among Measurement
Errors
var_a
var_a
var_p
eps1
eps2
1
1
anomia67
1
eps3
powles67
path_p
67
alienation
1
anomia71
powles71
1
path_p
71
alienation
ses
1
educatio
eps4
1
1
zeta1
var_p
SEI
1
1
delta1
delta2
Example 6: Model C
Exploratory analysis
W heaton (1977)
Model Specification
1
zeta2
Covariances among the Errors in
the Equations
GPA
academic
1
error1
height
weight
attract
1
error2
rating
Example 7
A nonrecursive model
Felson and Bohrnstedt (1979)
(Female subjects)
Over Identified Model
No. of Variances/Covariances = 21
No. of Parameters Estimated = 20
DF = 1
Class Exercise
• Create a new model:
– From the menu choose File/New
• Specify the Data file:
– Choose File/Data Files
– Browse to the tutorial folder. The path is:
• C:\Program Files\Amos 6\Examples
• In the Files of type list select SPSS
• Select Fels_mal.sav
• Estimate the Parameters of the different models
and compare their fit statistics