Structural equation
modeling
Advanced topics
Rex B Kline
Concordia
QICSS Set C
C1
Invariance
Multiple samples
o 1 model
o 2 samples
o Simultaneous analysis
C2
D1
1
X1
Y1
D2
1
X2
Y2
Group as moderator
Group model
C3
Multiple samples
o Test for differences
o Equal unstandardized
o Relative model fit
C4
CFA
o Measurement invariance
o Same factor model
o Groups, times, or forms
C5
Types
o
o
o
o
o
Dimensional
Configural
Weak (Metric)
Strong (Scalar)
Strict (Error)
C6
Configural
o Same model
o No other constraints
o Different scoring systems
C7
Weak
o Assumes configural
o Equal pattern coefficients
o Same scoring system
C8
Strong
o Assumes weak
o Equal intercepts or thresholds
o Same level of responding
C9
Strict
o Assumes strong
o Same error structure
o Identical measurement
C10
Strict
o
o
DeShon, R. P. (2004). Measures are not invariant across
groups without error variance homogeneity.
Psychology Science, 46, 137149.
Wu, A. D., Li, Z., & Zumbo, B. D. (2007). Decoding the
meaning of factorial Invariance and updating the
practice of multi-group confirmatory factor analysis:
A demonstration with TIMSS data. Practical
Assessment Research & Evaluation, 12(3). Retrieved
from http://pareonline.net/pdf/v12n3.pdf
C11
Strategy
o Model trimming
o Free baseline method:
Configural Weak Strong Strict
C12
Strategy
o Model building
o Constrained baseline method:
Strong ?
C13
Partial invariance
o Example: Weak rejected
o Some, not all
o DIF (1)
C14
Partial invariance
o Weak rejected:
1. Different meanings
2. Extreme response style
C15
Partial invariance
o
Cheung, G. W., & Rensvold, R. B. (2000).
Assessing extreme and
acquiescence response sets in crosscultural research using structural
equations modeling. Journal of
Cross-Cultural Psychology, 31, 187
212.
C16
Partial invariance
o Example: Strong rejected
o Some, not all
o DIF (2)
C17
Partial invariance
o Strong rejected
o Differential additive response
style
o Gender, cultural, procedural
C18
Partial invariance
o
Steinmetz, H. (2011). Analyzing
observed composite differences
across groups: Is partial
measurement invariance enough?
Methodology, 9, 112.
C19
Testing
2
D
o and N
o Low power unless n = 400
o Very large n
C20
Testing
o
Cheung, G. W., & Rensvold, R. B. (2002).
Evaluating goodness-of-fit indexes
for testing measurement invariance.
Structural Equation Modeling, 9, 233
255.
C21
Testing
o CFI .01 rule
2
M
2
B
max ( dfM , 0)
o CFI = 1
max ( dfB , 0)
C22
Testing
o
Chen, F. F. (2007). Sensitivity of
goodness of fit indexes to lack of
measurement invariance. Structural
Equation Modeling, 14, 464504.
More complicated CFI, RMSEA rules
C23
Testing
o Very large N
o Effect size
o Meaningful difference?
C24
Example
o
Millsap, R. E., & Olivera-Aguilar, M.
(2012). Investigating measurement
invariance using confirmatory factor
analysis. In R. H. Hoyle (Ed.),
Handbook of structural equation
modeling (pp. 380392). New York:
Guilford Press.
C25
Extensions
o Hispanic with adolescents
o n1 = 193 English
o n2 = 257 Spanish
C26
English
Item
fc1
fc4
fc5
fc6
fc9
SD
fc1
.381
.599
.416
.601
2.280
.887
fc4
.227
.393
.445
.404
1.518
.925
fc5
.400
.322
.476
.661
2.052
.972
fc6
.324
.330
.354
.519
1.689
1.014
fc9
.473
.370
.486
.540
1.684
.901
2.113
1.175
1.708
SD
1.034
.597
.836
Spanish
1.366 1.319
.785
.701
Note. English-speaking (above diagonal; n1 = 193), Spanish-speaking
(below diagonal; n2 = 257).
C27
E4
E1
1
fc1
fc4
E5
1
fc5
E6
1
fc6
E9
1
fc9
Conflict
C28
Identification
o
Little, T. D., Slegers, D. W., & Card, N. A.
(2006). A non-arbitrary method of
identifying and scaling latent
variables in SEM and MACS models.
Structural Equation Modeling, 13, 59
72.
C29
Reference-group method
E4
E1
1
fc1
fc4
fc5
E5
E6
E9
fc6
fc9
Conflict
C30
Marker-variable method
E4
E1
1
fc1
fc4
E5
1
fc5
E6
1
fc6
E9
1
fc9
1
0
Conflict
C31
Effects-coding method, loadings
E4
E1
1
fc4
fc1
1
E5
fc5
2
E6
1
fc6
3
E9
1
fc9
5
Conflict
C32
1 + 2 + 3 + 4 + 5
=1
5
1 = 5 2 3 4 5
2 = 5 1 3 4 5
3 = 5 1 2 4 5
4 = 5 1 2 3 5
5 = 5 1 2 3 4
C33
Effects-coding method, intercepts
E4
E1
1
fc1
E5
fc4
E6
fc5
fc6
E9
1
fc9
Conflict
C34
1 + 2 + 3 + 4 + 5
=0
5
1 = 0 2 3 4 5
2 = 0 1 3 4 5
3 = 0 1 2 4 5
4 = 0 1 2 3 5
5 = 0 1 2 3 4
C35
Models
1.
2.
3.
4.
5.
6.
Configural (M1)
M1 with error correlation (M2)
M2 with full metric (M3)
M3 with full error (M4)
M3 with full scalar (M5)
M3 with partial scalar (M6)
C36
Model 1 (No)
C37
Model 2 (Yes)
C38
Model 3 (Yes)
C39
Model 4 (No)
C40
Model 5 (No)
C41
Model 6 (Yes)
C42
Model
M2
dfM
D2
dfD Comparison RMSEA
90% CI
1 (N)
15.36
10
.049
0 .095
2 (Y)
7.85
7.51**
2 vs. 1
0 .068
3 (Y)
13.36
13
5.51*
3 vs. 2
.011
0 .068
4 (N)
87.57**
18
74.21**
4 vs. 3
.131
.104 .159
5 (N)
22.72
17
9.36
5 vs. 3
.039
0 .076
6 (Y)
13.46
15
.10
6 vs. 3
0 .057
C43
Model 1
E4
E1
1
fc1
fc4
E5
1
fc5
E6
1
fc6
E9
1
fc9
Conflict
Observations
v (v + 3)/2 no. groups
5(8)/2 2 = 40
C44
Parameters
Variances: (5 + 1) 2 = 12
Loadings: (4) 2 = 8
Intercepts = (4) 2 = 8
Means = 2
dfM = 40 30 = 10
C45
C46
title:
dillman data
model 1
configural invariance
data:
file is dillman.dat;
type is means std corr;
nobservations = 193 257;
ngroups = 2;
! group 1 is english, group 2 is spanish
variable:
names = fc1 fc4 fc5 fc6 fc9;
analysis:
type = general; estimator = ml;
C47
model:
! names group 1 factor loadings:
Conflict by fc1* (g1_load1)
fc4 (g1_load2)
fc5 (g1_load3)
fc6 (g1_load4)
fc9 (g1_load5);
! names group 1 intercepts:
[fc1] (g1_int1);
[fc4] (g1_int2);
[fc5] (g1_int3);
[fc6] (g1_int4);
[fc9] (g1_int5);
model g1:
! group 1 factor mean is free parameter:
[Conflict];
C48
model g2:
! names group 2 factor loadings
! separate loadings estimated in group 2:
Conflict by fc1* (g2_load1)
fc4 (g2_load2)
fc5 (g2_load3)
fc6 (g2_load4)
fc9 (g2_load5);
! names group 2 factor loadings
! separate intercepts estimated in group 2:
[fc1] (g2_int1);
[fc4] (g2_int2);
[fc5] (g2_int3);
[fc6] (g2_int4);
[fc9] (g2_int5);
C49
! separate factor mean, variance in group 2:
Conflict;
[Conflict];
! by default, measurement errors are freely
! estimated in each group
model constraint:
! effects-coding method for scaling
! and identifying factor
! average loading constrained to 1.0 and
! average intercept constrained to 0
C50
! group 1:
g1_load1 = 5 - g1_load2 - g1_load3 - g1_load4 - g1_load5;
g1_int1 = -g1_int2 - g1_int3 - g1_int4 - g1_int5;
! group 2:
g2_load1 = 5 - g2_load2 - g2_load3 - g2_load4 - g2_load5;
g2_int1 = -g2_int2 - g2_int3 - g2_int4 - g2_int5;
output: samp stdyx res;
C51
THE MODEL ESTIMATION TERMINATED NORMALLY
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
15.363
10
0.1194
Chi-Square Contributions From Each Group
G1
G2
8.991
6.372
C52
CFI/TLI
CFI
TLI
0.991
0.983
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.049
0.000
0.095
SRMR (Standardized Root Mean Square Residual)
Value
0.026
C53
RESIDUAL OUTPUT
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G1
Model Estimated Covariances/Correlations/Residual Correlations
FC1
FC4
FC5
FC6
________
________
________
________
FC1
0.783
FC4
0.312
0.851
FC5
0.499
0.373
0.940
FC6
0.406
0.304
0.487
1.023
FC9
0.477
0.357
0.572
0.465
Residuals for Covariances/Correlations/Residual Correlations
FC1
FC4
FC5
FC6
________
________
________
________
FC1
0.000
FC4
-0.001
0.000
FC5
0.014
-0.022
0.000
FC6
-0.034
0.112
-0.020
0.000
FC9
0.001
-0.022
0.004
0.007
FC9
________
0.808
FC9
________
0.000
C54
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G1
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
0.014
FC4
-0.019
0.000
FC5
0.843
-0.924
0.000
FC6
-1.365
2.466
-0.941
0.000
FC9
0.055
-1.151
0.396
0.324
999.000
C55
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G2
Model Estimated Covariances/Correlations/Residual Correlations
FC1
FC4
FC5
FC6
________
________
________
________
FC1
1.065
FC4
0.164
0.355
FC5
0.298
0.141
0.696
FC6
0.293
0.139
0.252
0.614
FC9
0.338
0.160
0.292
0.287
Residuals for Covariances/Correlations/Residual Correlations
FC1
FC4
FC5
FC6
________
________
________
________
FC1
0.000
FC4
-0.024
0.000
FC5
0.047
0.019
0.000
FC6
-0.031
0.015
-0.021
0.000
FC9
0.003
-0.006
-0.008
0.009
FC9
________
0.489
FC9
________
0.000
C56
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G2
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
999.000
FC4
-1.011
0.000
FC5
1.548
0.978
0.000
FC6
-1.301
0.884
-1.195
0.000
FC9
0.305
-0.880
-1.265
1.183
0.000
C57
PAGE :
TITLE:
EQS
Group 1: English
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V4, V2
V4, V1
V5, V2
V3, V2
V4, V3
V3, V1
V5, V4
V5, V3
V5, V1
V2, V1
ESTIMATE
-------.119
-.038
-.026
-.024
-.020
.017
.007
.005
.001
-.001
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V4, V4
V2, V2
V3, V3
V5, V5
V999,V4
V999,V5
V999,V2
V999,V3
V999,V1
V1, V1
ESTIMATE
-------.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
C58
PAGE : 12
TITLE:
EQS
Group 2: Spanish
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V3, V1
V2, V1
V3, V2
V4, V1
V4, V2
V4, V3
V5, V4
V5, V2
V5, V3
V5, V1
ESTIMATE
-------.054
-.039
.038
-.038
.033
-.032
.017
-.015
-.014
.005
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V5, V5
V4, V4
V3, V3
V2, V2
V999,V5
V999,V4
V999,V3
V999,V2
V999,V1
V1, V1
ESTIMATE
-------.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
C59
Model 2 (Yes)
C60
title:
dillman-carpentier
model 2
configural invariance with error correlation
group 1, fc4 and fc6
...
model g1:
! group 1 factor mean is free parameter:
[Conflict];
! ** new to model 2 **
! error covariance group 1 only:
fc4 with fc6;
C61
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
7.850
9
0.5493
Chi-Square Contributions From Each Group
G1
G2
1.478
6.372
CFI
TLI
1.000
1.004
CFI/TLI
C62
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.000
0.000
0.068
SRMR (Standardized Root Mean Square Residual)
Value
0.018
C63
RESIDUAL OUTPUT
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G1
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
999.000
FC4
0.487
0.032
FC5
0.617
-0.195
999.000
FC6
-0.847
0.020
-0.321
0.010
FC9
-0.401
-0.257
-0.437
0.916
999.000
C64
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G2
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
999.000
FC4
-1.011
0.000
FC5
1.547
0.977
0.000
FC6
-1.302
0.884
-1.195
999.000
FC9
0.305
-0.880
-1.265
1.184
999.000
C65
PAGE :
TITLE:
EQS
Group 1: English
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V4, V1
V5, V4
V2, V1
V3, V1
V4, V3
V5, V2
V3, V2
V5, V1
V5, V3
V5, V5
ESTIMATE
--------.025
.021
.018
.011
-.007
-.006
-.005
-.005
-.003
.000
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V4, V2
V4, V4
V3, V3
V2, V2
V1, V1
V999,V5
V999,V2
V999,V4
V999,V1
V999,V3
ESTIMATE
-------.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
C66
PAGE : 14
TITLE:
EQS
Group 2: Spanish
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V3, V1
V2, V1
V3, V2
V4, V1
V4, V2
V4, V3
V5, V4
V5, V2
V5, V3
V5, V1
ESTIMATE
-------.054
-.039
.038
-.038
.033
-.032
.017
-.015
-.014
.005
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V5, V5
V4, V4
V3, V3
V2, V2
V999,V5
V999,V4
V999,V3
V999,V2
V999,V1
V1, V1
ESTIMATE
-------.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
C67
Model 3 (Yes)
C68
title:
dillman-carpentier
model 3
metric invariance with error correlation
group 1, fc4 and fc6
...
C69
model g2:
! names group 2 factor loadings
! ** new to model 3 **
! pairwise equality constraints imposed on loadings
! by commenting out the following group 2 syntax:
! Conflict by fc1* (g2_load1)
!
fc4 (g2_load2)
!
fc5 (g2_load3)
!
fc6 (g2_load4)
!
fc9 (g2_load5);
...
C70
model constraint:
! effects-coding method for scaling
! and identifying factor
! average loading constrained to 1.0 and
! average intercept constrained to 0
! group 1:
g1_load1 = 5 - g1_load2 - g1_load3 - g1_load4 - g1_load5;
g1_int1 = -g1_int2 - g1_int3 - g1_int4 - g1_int5;
! group 2:
! ** new to model 3 **
! original group 2 constraint on loadings not needed
! and is commented out:
! g2_load1 = 5 - g2_load2 - g2_load3 - g2_load4 - g2_load5;
C71
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
13.361
13
0.4203
Chi-Square Contributions From Each Group
G1
G2
3.617
9.744
CFI
TLI
0.999
0.999
CFI/TLI
C72
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.011
0.000
0.068
SRMR (Standardized Root Mean Square Residual)
Value
0.036
C73
RESIDUAL OUTPUT
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G1
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
999.000
FC4
0.744
0.850
FC5
0.404
1.019
1.079
FC6
-1.764
0.565
0.028
-0.626
FC9
-3.828
0.675
0.839
-0.074
-0.274
C74
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED)
FOR G2
Standardized Residuals (z-scores) for Covariances/Correlations/Residual Corr
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
FC1
1.205
FC4
-0.604
-1.958
FC5
1.356
-0.535
-2.635
FC6
0.402
0.354
-1.347
0.688
FC9
1.511
-1.508
-2.435
1.302
0.228
C75
PAGE :
TITLE:
EQS
Group 1: English
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V4, V1
V3, V2
V1, V1
V3, V3
V5, V1
V2, V1
V2, V2
V5, V2
V5, V3
V4, V4
ESTIMATE
--------.065
.057
-.048
.047
-.043
.041
.041
.035
.026
-.023
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V4, V2
V3, V1
V5, V5
V5, V4
V4, V3
V999,V5
V999,V4
V999,V1
V999,V2
V999,V3
ESTIMATE
-------.021
.013
-.007
-.003
.001
.000
.000
.000
.000
.000
C76
PAGE : 14
TITLE:
EQS
Group 2: Spanish
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V5, V1
V3, V1
V3, V3
V1, V1
V5, V3
V4, V3
V5, V2
V5, V4
V2, V2
V2, V1
ESTIMATE
-------.068
.067
-.057
.055
-.055
-.052
-.048
.047
-.035
-.031
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V4, V4
V3, V2
V4, V1
V4, V2
V5, V5
V999,V3
V999,V2
V999,V1
V999,V4
V999,V5
ESTIMATE
-------.026
-.022
.020
.016
.006
.000
.000
.000
.000
.000
C77
Model 4 (No)
C78
title:
dillman-carpentier
model 4
metric invariance with error correlation
group 1, fc4 and fc6
equality of error variances
...
C79
model:
! names group 1 factor loadings:
Conflict by fc1* (g1_load1)
fc4 (g1_load2)
fc5 (g1_load3)
fc6 (g1_load4)
fc9 (g1_load5);
! names group 1 intercepts:
[fc1] (g1_int1);
[fc4] (g1_int2);
[fc5] (g1_int3);
[fc6] (g1_int4);
[fc9] (g1_int5);
! ** new to model 4 **
! names group 1 error variances:
fc1 (g1_err1);
fc4 (g1_err2);
fc5 (g1_err3);
fc6 (g1_err4);
fc9 (g1_err5);
C80
model g2:
! names group 2 factor loadings
! pairwise equality constraints imposed on loadings
! by commenting out the following group 2 syntax:
! Conflict by fc1* (g2_load1)
!
fc4 (g2_load2)
!
fc5 (g2_load3)
!
fc6 (g2_load4)
!
fc9 (g2_load5);
! names group 2 factor loadings
! separate intercepts estimated in group 2:
[fc1] (g2_int1);
[fc4] (g2_int2);
[fc5] (g2_int3);
[fc6] (g2_int4);
[fc9] (g2_int5);
! ** new to model 4 **
! names group 2 error variances:
fc1 (g2_err1);
fc4 (g2_err2);
fc5 (g2_err3);
fc6 (g2_err4);
fc9 (g2_err5);
C81
model constraint:
! effects-coding method for scaling
! and identifying factor
! average loading constrained to 1.0 and
! average intercept constrained to 0
! group 1:
g1_load1 = 5 - g1_load2 - g1_load3 - g1_load4 - g1_load5;
g1_int1 = -g1_int2 - g1_int3 - g1_int4 - g1_int5;
! group 2:
! original group 2 constraint on loadings not needed
! and is commented out:
! g2_load1 = 5 - g2_load2 - g2_load3 - g2_load4 - g2_load5;
g2_int1 = -g2_int2 - g2_int3 - g2_int4 - g2_int5;
! ** new to model 4 **
! constrain error variances:
g1_err1 = g2_err1;
g1_err2 = g2_err2;
g1_err3 = g2_err3;
g1_err4 = g2_err4;
g1_err5 = g2_err5;
C82
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
87.571
18
0.0000
Chi-Square Contributions From Each Group
G1
G2
42.617
44.954
CFI
TLI
0.888
0.876
CFI/TLI
C83
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.131
0.104
0.159
SRMR (Standardized Root Mean Square Residual)
Value
0.122
C84
C85
Model 3 (Yes)
C86
Model 3
MODEL RESULTS
Two-Tailed
P-Value
Estimate
S.E.
Est./S.E.
0.355
0.651
0.353
0.648
0.249
0.047
0.069
0.049
0.073
0.041
7.592
9.397
7.207
8.825
6.021
0.000
0.000
0.000
0.000
0.000
0.682
0.808
0.572
0.685
0.384
0.043
0.036
0.047
0.043
0.051
15.807
22.353
12.283
15.951
7.570
0.000
0.000
0.000
0.000
0.000
Group G1
Residual Variances
FC1
FC4
FC5
FC6
FC9
Group G2
Residual Variances
FC1
FC4
FC5
FC6
FC9
C87
Model 5 (metric + scalar invariance)
C88
title:
dillman-carpentier
model 5
metric invariance with error correlation
group 1, fc4 and fc6
scalar invariance
...
C89
model g2:
! names group 2 factor loadings
! pairwise equality constraints imposed on loadings
! by commenting out the following group 2 syntax:
! Conflict by fc1* (g2_load1)
!
fc4 (g2_load2)
!
fc5 (g2_load3)
!
fc6 (g2_load4)
!
fc9 (g2_load5);
! names group 2 factor loadings
! separate intercepts estimated in group 2:
! ** new to model 5 **
! comment out naming of constaints in group 2
! now pairwise constrained to equal group 1 values:
![fc1] (g2_int1);
![fc4] (g2_int2);
![fc5] (g2_int3);
![fc6] (g2_int4);
![fc9] (g2_int5);
C90
model constraint:
! effects-coding method for scaling
! and identifying factor
! average loading constrained to 1.0 and
! average intercept constrained to 0
! group 1:
g1_load1 = 5 - g1_load2 - g1_load3 - g1_load4 - g1_load5;
g1_int1 = -g1_int2 - g1_int3 - g1_int4 - g1_int5;
! group 2:
! original group 2 constraint on loadings not needed
! and is commented out:
! g2_load1 = 5 - g2_load2 - g2_load3 - g2_load4 - g2_load5;
! ** new to model 5 **
! original group 2 constraint on loadings not needed
! g2_int1 = -g2_int2 - g2_int3 - g2_int4 - g2_int5;
C91
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
22.717
17
0.1587
Chi-Square Contributions From Each Group
G1
G2
7.779
14.938
CFI/TLI
CFI
TLI
0.991
0.989
C92
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.039
0.000
0.076
SRMR (Standardized Root Mean Square Residual)
Value
0.046
C93
RESIDUAL OUTPUT
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED) FOR G1
Model Estimated Means/Intercepts/Thresholds
FC1
FC4
FC5
FC6
________
________
________
________
2.334
1.410
2.052
1.664
FC9
________
1.676
Residuals for Means/Intercepts/Thresholds
FC1
FC4
FC5
________
________
________
-0.054
0.108
0.000
FC9
________
0.008
FC6
________
0.025
Standardized Residuals (z-scores) for Means/Intercepts/Thresholds
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
-2.418
2.272
-0.005
0.603
0.437
C94
ESTIMATED MODEL AND RESIDUALS (OBSERVED - ESTIMATED) FOR G2
Model Estimated Means/Intercepts/Thresholds
FC1
FC4
FC5
FC6
________
________
________
________
2.028
1.209
1.708
1.367
FC9
________
1.323
Residuals for Means/Intercepts/Thresholds
FC1
FC4
FC5
________
________
________
0.085
-0.034
0.000
FC9
________
-0.004
FC6
________
-0.001
Standardized Residuals (z-scores) for Means/Intercepts/Thresholds
FC1
FC4
FC5
FC6
FC9
________
________
________
________
________
2.060
-2.788
0.007
-0.040
-0.380
C95
PAGE :
TITLE:
EQS
Group 1: English
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V999,V2
V999,V1
V4, V1
V3, V3
V3, V2
V2, V1
V4, V4
V3, V1
V5, V1
V999,V4
ESTIMATE
-------.117
-.061
-.054
.042
.035
.032
-.032
.031
-.027
.024
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V1, V1
V5, V3
V5, V5
V5, V4
V999,V5
V5, V2
V4, V3
V2, V2
V4, V2
V999,V3
ESTIMATE
--------.021
.016
-.015
-.015
.009
.008
-.007
.006
-.001
.000
C96
PAGE : 14
TITLE:
EQS
Group 2: Spanish
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MAXIMUM LIKELIHOOD SOLUTION (NORMAL DISTRIBUTION THEORY)
LARGEST STANDARDIZED RESIDUALS:
NO.
--1
2
3
4
5
6
7
8
9
10
PARAMETER
--------V5, V1
V3, V1
V999,V1
V5, V2
V1, V1
V2, V2
V999,V2
V3, V3
V5, V3
V4, V3
ESTIMATE
-------.088
.085
.082
-.068
.062
-.061
-.056
-.052
-.050
-.049
NO.
--11
12
13
14
15
16
17
18
19
20
PARAMETER
--------V5, V4
V3, V2
V4, V1
V2, V1
V4, V4
V5, V5
V999,V5
V999,V4
V999,V3
V4, V2
ESTIMATE
-------.047
-.037
.034
-.031
.027
.007
-.006
-.001
.000
.000
C97
Model 6 (metric + partial scalar)
C98
title:
dillman-carpentier
model 6
metric invariance with error correlation
group 1, fc4 and fc6
partial scalar invariance
...
C99
model g2:
! names group 2 factor loadings
! pairwise equality constraints imposed on loadings
! by commenting out the following group 2 syntax:
! Conflict by fc1* (g2_load1)
!
fc4 (g2_load2)
!
fc5 (g2_load3)
!
fc6 (g2_load4)
!
fc9 (g2_load5);
! names group 2 factor loadings
! separate intercepts estimated in group 2:
! ** new to model 6 **
! comment out naming of constraints in group 2
! for only 3 indicators:
[fc1] (g2_int1);
[fc4] (g2_int2);
![fc5] (g2_int3);
![fc6] (g2_int4);
![fc9] (g2_int5);
C100
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value
Degrees of Freedom
P-Value
13.462
15
0.5667
Chi-Square Contributions From Each Group
G1
G2
3.808
9.654
CFI
TLI
1.000
1.003
CFI/TLI
C101
RMSEA (Root Mean Square Error Of Approximation)
Estimate
90 Percent C.I.
0.000
0.000
0.057
SRMR (Standardized Root Mean Square Residual)
Value
0.036
C102
MODEL RESULTS
Two-Tailed
P-Value
Estimate
S.E.
Est./S.E.
CONFLICT BY
FC1
FC4
FC5
FC6
FC9
1.062
0.635
1.144
0.988
1.171
0.059
0.055
0.053
0.056
0.049
18.033
11.570
21.441
17.706
23.847
0.000
0.000
0.000
0.000
0.000
FC4
FC6
0.139
0.052
2.657
0.008
1.062
0.635
1.144
0.988
1.171
0.059
0.055
0.053
0.056
0.049
18.033
11.570
21.441
17.706
23.847
0.000
0.000
0.000
0.000
0.000
Group G1
WITH
Group G2
CONFLICT BY
FC1
FC4
FC5
FC6
FC9
C103
MODEL RESULTS
Two-Tailed
P-Value
Estimate
S.E.
Est./S.E.
0.355
0.651
0.354
0.648
0.249
0.047
0.069
0.049
0.073
0.041
7.592
9.395
7.262
8.821
6.040
0.000
0.000
0.000
0.000
0.000
0.741
0.273
0.427
0.370
0.165
0.071
0.026
0.045
0.038
0.027
10.438
10.400
9.429
9.666
6.173
0.000
0.000
0.000
0.000
0.000
Group G1
Residual Variances
FC1
FC4
FC5
FC6
FC9
Group G2
Residual Variances
FC1
FC4
FC5
FC6
FC9
C104
MODEL RESULTS
Two-Tailed
P-Value
Estimate
S.E.
Est./S.E.
Means
CONFLICT
1.843
0.051
36.087
0.000
Intercepts
FC1
FC4
FC5
FC6
FC9
0.323
0.346
-0.051
-0.144
-0.474
0.115
0.113
0.095
0.097
0.087
2.797
3.062
-0.531
-1.476
-5.427
0.005
0.002
0.595
0.140
0.000
Means
CONFLICT
1.532
0.039
39.431
0.000
Intercepts
FC1
FC4
FC5
FC6
FC9
0.486
0.202
-0.051
-0.144
-0.474
0.105
0.089
0.095
0.097
0.087
4.641
2.263
-0.531
-1.476
-5.427
0.000
0.024
0.595
0.140
0.000
Group G1
Group G2
C105
MODEL RESULTS
Two-Tailed
P-Value
Estimate
S.E.
Est./S.E.
Means
CONFLICT
1.843
0.051
36.087
0.000
Variances
CONFLICT
0.412
0.051
8.151
0.000
Means
CONFLICT
1.532
0.039
39.431
0.000
Variances
CONFLICT
0.235
0.027
8.759
0.000
Group G1
Group G2
1.843 1.532 .311
=
= .48
.642
.412
1.843 1.532 .311
=
= .64
.484
.235
C106
Summary
o Same scoring model (metric)
o But different precisions (error)
o Different levels (thresholds)
C107
Ordinal data
o
Millsap, R. E. (2011). Statistical approaches to
measurement invariance. New York:
Routledge.
Millsap, R. E., & YunTein, J. (2004). Assessing
factorial invariance in ordered-categorical
measures. Multivariate Behavioral
Research, 39, 479515.
C108
Ordinal data
o
Edwards, M. C., Wirth, R. J., Houts, C. R., & Xi, N.
(2012). Categorical data in the structural
equation modeling framework. In R. Hoyle (Ed.),
Handbook of structural equation modeling (pp.
195208). New York: Guilford.
Finney, S. J., & DiStefano, C. (2013). Nonnormal and
categorical data in structural equation modeling.
In G. R. Hancock & R. O. Mueller (Eds.), Structural
equation modeling: A second course (2nd ed.)
(pp. 269314). Charlotte, NC: IAP.
C109
X1
X2
1113
3133
4143
5153
X1*
E3
E2
X2*
E4
E5
X5*
X4*
X3*
2
5
X5
2123
E1
X4
X3
Depression
C110
Ordinal data
o Threshold structure
o Indicators are LRVs
o Mean structure
C111
Ordinal data
o Scaling (parameterization):
Delta
Theta
C112
Ordinal data
o Delta:
var (LRV) = 1.0
Polychoric metric
C113
Ordinal data
o Theta:
Residual var (LRV) = 1.0
Probit regression
C114
Ordinal data
o Invariance testing:
Theta > delta
Res. var = 1.0 in reference
Freely estimated in others
C115
Ordinal data
o Configural
o Weak
o Strong
o Strict
C116
Ordinal data
o IRT
o ICC
o SEM and IRT
C117
1.0
Probability of Correct
Response
.9
.8
.7
.6
ICC
.5
Difficulty
.4
Tangent line
.3
.2
.1
0
3.0
2.0
1.0
1.0
2.0
3.0
Latent Ability ()
C118
APPENDIX
MULTIPLE-SAMPLES PA
C119
MG path analysis
o
Lynam, D. R., Moffitt, T., & StouthamerLoeber, M. (1993). Explaining the
relation between IQ and
delinquency: Class, race, test
motivation, or self-control? Journal
of Abnormal Psychology, 102, 187
196.
C120
MG path analysis
o Adolescents, low SES
o n1 = 214 African-American
o n2 = 181 white
C121
DAc
Social
Class
Achievement
Motivation
DDe
1
Verbal
Ability
Delinquency
C122
AfricanAmerican
Variable
1.
2.
3.
4.
5.
Social Class
Motivation
Verbal Ability
Achievement
Delinquency
White
M
SD
SD
.25
.37
.27
.11
.08
.40
.28
.20
.28
.30
.61
.31
.05
.21
.50
.21
.11
.17
.26
.33
31.96
.01
93.76
2.51
1.40
10.58
1.35
13.62
.79
1.63
34.64
11.53
.05
1.32
104.18
16.32
2.88
.96
1.22
1.45
Note. African-American (above diagonal), white (below diagonal). Means are
reported but not analyzed.
C123
DAc
Social
Class
Achievement
Motivation
DDe
1
Verbal
Ability
Delinquency
Observations
v (v + 1)/2 no. groups
5(6)/2 2 = 30
C124
DAc
Social
Class
Achievement
Motivation
DDe
1
Verbal
Ability
Delinquency
Parameters
Variances: (3 + 2) 2 = 10
Covariances = 3 2 = 6
Direct effects = 7
dfM = 30 (10 + 6 + 7) = 7
C125
/title
eqs
lynam et al. (1993) white
/specifications
variables = 5; cases = 181;
matrix = cor ; groups=2;
/labels
v1 = soc_cls; v2 = test_eff; v3 = verbal;
v4 = achieve; v5 = delinq;
/equations
v4 = *v1 +*v2 + *v3 + e4;
v5 = *v1 +*v2 + *v3 + *v4 + e5;
/variances
v1-v3 = *; e4-e5 = *;
/covariances
v1,v2 = *; v1,v3 = *; v2,v3 = *;
C126
/matrix
1.00
.25 1.00
.37 .40 1.00
.27 .28 .61 1.00
-.11 -.20 -.31 -.21 1.00
/standard deviations
11.53 1.32 16.32 .96 1.45
/end
C127
/title
african-american
/specifications
variables = 5; cases = 214; matrix = cor;
/labels
v1 = soc_cls; v2 = test_eff; v3 = verbal;
v4 = achieve; v5 = delinq;
/equations
v4 = *v1 + *v2 + *v3 +e4;
v5 = *v1 + *v2 + *v3 + *v4 + e5;
/variances
v1-v3 = *; e4-e5 = *;
/covariances
v1,v2 = *; v1,v3 = *; v2,v3 = *;
C128
/matrix
1.00
.08 1.00
.28 .30 1.00
.05 .21 .50 1.00
-.11 -.17 -.26 -.33 1.00
/standard deviations
10.58 1.35 13.62 .79 1.63
C129
/print
fit = all;
/constraints
(1,v4,v1) =
(1,v4,v2) =
(1,v4,v3) =
(1,v5,v1) =
(1,v5,v2) =
(1,v5,v3) =
(1,v5,v4) =
/lmtest
/end
(2,v4,v1);
(2,v4,v2);
(2,v4,v3);
(2,v5,v1);
(2,v5,v2);
(2,v5,v3);
(2,v5,v4);
V1
Social
Class
DAc
1
V4
Achievement
V2
Motivation
DDe
V5
V3
Verbal
Ability
Delinquency
C130
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP 1
STANDARDIZED RESIDUAL MATRIX:
SOC_CLS
TEST_EFF
VERBAL
V1
V2
V3
SOC_CLS V1
.000
TEST_EFF V2
.000
.000
VERBAL
V3
.000
.000
.000
ACHIEVE V4
.086
.020
.060
DELINQ
V5
.030
.016
.016
ACHIEVE
V4
.061
.080
AVERAGE ABSOLUTE STANDARDIZED RESIDUAL =
AVERAGE OFF-DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL =
LARGEST STANDARDIZED RESIDUALS:
NO.
PARAMETER
ESTIMATE
-----------------1
V4, V1
.086
2
V5, V4
.080
3
V4, V4
.061
4
V4, V3
.060
5
V5, V5
-.044
6
V5, V1
.030
7
V4, V2
.020
8
V5, V2
.016
9
V5, V3
.016
10
V3, V3
.000
NO.
--11
12
13
14
15
PARAMETER
--------V3, V2
V3, V1
V2, V2
V2, V1
V1, V1
DELINQ
V5
-.044
.0274
.0307
ESTIMATE
-------.000
.000
.000
.000
.000
C131
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP 2
STANDARDIZED RESIDUAL MATRIX:
SOC_CLS
TEST_EFF
VERBAL
V1
V2
V3
SOC_CLS V1
.000
TEST_EFF V2
.000
.000
VERBAL
V3
.000
.000
.000
ACHIEVE V4
-.077
-.016
-.058
DELINQ
V5
-.035
-.015
-.021
ACHIEVE
V4
-.061
-.099
AVERAGE ABSOLUTE STANDARDIZED RESIDUAL =
AVERAGE OFF-DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL =
LARGEST STANDARDIZED RESIDUALS:
NO.
PARAMETER
ESTIMATE
-----------------1
V5, V4
-.099
2
3
4
5
6
7
8
9
10
V4,
V4,
V4,
V5,
V5,
V5,
V4,
V5,
V3,
V1
V4
V3
V5
V1
V3
V2
V2
V3
-.077
-.061
-.058
.038
-.035
-.021
-.016
-.015
.000
NO.
--11
PARAMETER
--------V3, V2
12
13
14
15
V3,
V2,
V2,
V1,
V1
V2
V1
V1
DELINQ
V5
.038
.0280
.0322
ESTIMATE
-------.000
.000
.000
.000
.000
C132
GOODNESS OF FIT SUMMARY FOR METHOD = ML
INDEPENDENCE MODEL CHI-SQUARE
INDEPENDENCE AIC =
MODEL AIC =
257.348
-2.324
297.348 ON
INDEPENDENCE CAIC =
MODEL CAIC =
20 DEGREES OF FREEDOM
157.770
-37.176
CHI-SQUARE =
11.676 BASED ON
7 DEGREES OF FREEDOM
PROBABILITY VALUE FOR THE CHI-SQUARE STATISTIC IS
.11172
FIT INDICES
----------BENTLER-BONETT
NORMED FIT INDEX =
.961
BENTLER-BONETT NON-NORMED FIT INDEX =
.952
COMPARATIVE FIT INDEX (CFI)
=
.983
BOLLEN'S
(IFI) FIT INDEX
=
.984
MCDONALD'S
(MFI) FIT INDEX
=
.994
JORESKOG-SORBOM'S GFI FIT INDEX
=
.988
JORESKOG-SORBOM'S AGFI FIT INDEX
=
.951
ROOT MEAN-SQUARE RESIDUAL (RMR)
=
.347
STANDARDIZED RMR
=
.041
ROOT MEAN-SQUARE ERROR OF APPROXIMATION (RMSEA)
90% CONFIDENCE INTERVAL OF RMSEA (
.000,
.058
.115)
C133
LAGRANGE MULTIPLIER TEST (FOR RELEASING CONSTRAINTS)
CONSTRAINTS TO BE RELEASED ARE:
CONSTRAINTS FROM GROUP
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
1
2
3
4
5
6
7
(1,V4,V1)-(2,V4,V1)=0;
(1,V4,V2)-(2,V4,V2)=0;
(1,V4,V3)-(2,V4,V3)=0;
(1,V5,V1)-(2,V5,V1)=0;
(1,V5,V2)-(2,V5,V2)=0;
(1,V5,V3)-(2,V5,V3)=0;
(1,V5,V4)-(2,V5,V4)=0;
UNIVARIATE TEST STATISTICS:
NO
-1
2
3
4
5
6
7
CONSTRAINT
----------CONSTR:
1
CONSTR:
2
CONSTR:
3
CONSTR:
4
CONSTR:
5
CONSTR:
6
CONSTR:
7
CHI-SQUARE
---------3.803
.182
1.985
.883
.144
.325
5.455
PROBABILITY
----------.051
.669
.159
.348
.704
.569
.020
C134
MG path analysis
o Second analysis:
Achieve Delinq free
dfM = 6
C135
Achievement Delinquency free
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
STANDARDIZED RESIDUAL MATRIX:
SOC_CLS
TEST_EFF
V1
V2
SOC_CLS V1
.000
TEST_EFF V2
.000
.000
VERBAL
V3
.000
.000
ACHIEVE V4
.086
.020
DELINQ
V5
.023
-.009
VERBAL
V3
.000
.060
-.036
ACHIEVE
V4
.061
-.019
AVERAGE ABSOLUTE STANDARDIZED RESIDUAL =
AVERAGE OFF-DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL =
LARGEST STANDARDIZED RESIDUALS:
NO.
PARAMETER
ESTIMATE
-----------------1
V4, V1
.086
2
V4, V4
.061
3
V4, V3
.060
4
V5, V3
-.036
5
V5, V1
.023
6
V4, V2
.020
7
V5, V4
-.019
8
V5, V5
.015
9
V5, V2
-.009
10
V3, V3
.000
NO.
--11
12
13
14
15
PARAMETER
--------V3, V2
V3, V1
V2, V2
V2, V1
V1, V1
DELINQ
V5
.015
.0219
.0252
ESTIMATE
-------.000
.000
.000
.000
.000
C136
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP 2
STANDARDIZED RESIDUAL MATRIX:
SOC_CLS
TEST_EFF
VERBAL
V1
V2
V3
SOC_CLS V1
.000
TEST_EFF V2
.000
.000
VERBAL
V3
.000
.000
.000
ACHIEVE V4
-.077
-.016
-.058
DELINQ
V5
-.010
.011
.049
ACHIEVE
V4
-.061
.026
AVERAGE ABSOLUTE STANDARDIZED RESIDUAL =
AVERAGE OFF-DIAGONAL ABSOLUTE STANDARDIZED RESIDUAL =
LARGEST STANDARDIZED RESIDUALS:
NO.
PARAMETER
ESTIMATE
-----------------1
V4, V1
-.077
2
V4, V4
-.061
3
V4, V3
-.058
4
V5, V3
.049
5
V5, V4
.026
6
7
8
9
10
V5,
V4,
V5,
V5,
V2,
V5
V2
V2
V1
V1
NO.
--11
12
13
14
15
PARAMETER
--------V3, V3
V3, V2
V3, V1
V2, V2
V1, V1
DELINQ
V5
-.019
.0218
.0247
ESTIMATE
-------.000
.000
.000
.000
.000
-.019
-.016
.011
-.010
.000
C137
GOODNESS OF FIT SUMMARY FOR METHOD = ML
INDEPENDENCE MODEL CHI-SQUARE
INDEPENDENCE AIC =
MODEL AIC =
257.348
-5.924
297.348 ON
INDEPENDENCE CAIC =
MODEL CAIC =
20 DEGREES OF FREEDOM
157.770
-35.797
CHI-SQUARE =
6.076 BASED ON
6 DEGREES OF FREEDOM
PROBABILITY VALUE FOR THE CHI-SQUARE STATISTIC IS
.41471
FIT INDICES
----------BENTLER-BONETT
NORMED FIT INDEX =
.980
BENTLER-BONETT NON-NORMED FIT INDEX =
.999
COMPARATIVE FIT INDEX (CFI)
=
1.000
BOLLEN'S
(IFI) FIT INDEX
=
1.000
MCDONALD'S
(MFI) FIT INDEX
=
1.000
JORESKOG-SORBOM'S GFI FIT INDEX
=
.994
JORESKOG-SORBOM'S AGFI FIT INDEX
=
.969
ROOT MEAN-SQUARE RESIDUAL (RMR)
=
.396
STANDARDIZED RMR
=
.034
ROOT MEAN-SQUARE ERROR OF APPROXIMATION (RMSEA)
90% CONFIDENCE INTERVAL OF RMSEA (
.000,
.008
.093)
C138
LAGRANGE MULTIPLIER TEST (FOR RELEASING CONSTRAINTS)
CONSTRAINTS TO BE RELEASED ARE:
CONSTRAINTS FROM GROUP
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
1
2
3
4
5
6
(1,V4,V1)-(2,V4,V1)=0;
(1,V4,V2)-(2,V4,V2)=0;
(1,V4,V3)-(2,V4,V3)=0;
(1,V5,V1)-(2,V5,V1)=0;
(1,V5,V2)-(2,V5,V2)=0;
(1,V5,V3)-(2,V5,V3)=0;
UNIVARIATE TEST STATISTICS:
NO
-1
2
3
4
5
6
CONSTRAINT
----------CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
CONSTR:
1
2
3
4
5
6
CHI-SQUARE
---------3.804
.182
1.985
.365
.030
.758
PROBABILITY
----------.051
.669
.159
.546
.861
.384
C139
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MEASUREMENT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.
DELINQ
=V5
-.088*V4
.125
-.705
.003*V1
.007
-.491
.099*V2
.059
-1.684
.017*V3
.006
-2.760@
STANDARDIZED SOLUTION:
DELINQ
=V5
-.057*V4
-.197*V3
+ 1.000 E5
R-SQUARED
.028*V1
.956 E5
.091*V2
.087
MULTIPLE POPULATION ANALYSIS, INFORMATION IN GROUP
MEASUREMENT EQUATIONS WITH STANDARD ERRORS AND TEST STATISTICS
STATISTICS SIGNIFICANT AT THE 5% LEVEL ARE MARKED WITH @.
DELINQ
=V5
-.492*V4
.139
-3.539@
.003*V1
.007
-.491
.099*V2
.059
-1.684
STANDARDIZED SOLUTION:
DELINQ
=V5
-.243*V4
-.144*V3
.017*V3
.006
-2.760@
+ 1.000 E5
R-SQUARED
+
.022*V1
.925 E5
.081*V2
.144
C140
