Example 29.22 Testing Equality of Covariance and Mean Matrices between Independent Groups
To make the specification of some standard MSTRUCT models for covariance and mean patterns more efficient, PROC CALIS defines
these standard models internally. You can use two options to invoke these builtin covariance and mean patterns easily. For
example, with the COVPATTERN= option, you can define the compound symmetry (COMPSYM) pattern for the covariance matrix or the equality of covariance matrices
between groups (EQCOVMAT). With the MEANPATTERN= option, you can define uniform means (UNIFORM) for the mean vector or the equality of mean vectors between groups (EQMEANVEC).
See the COVPATTERN= and the MEANPATTERN= options for details about the supported builtin covariance and mean patterns.
In Example 29.21, you test of the equality of covariance matrices between two groups. This example extends the application to the test of
equality of mean vectors between three independent groups by using the COVPATTERN= and MEANPATTERN= options together. The
“best” fit model for the data is explored. The following DATA steps define the covariance and mean matrices for the three independent
groups, respectively:
data g1(type=corr);
Input _type_ $ 18 _name_ $ 911 x1x9;
datalines;
corr x1 1. . . . . . . . .
corr x2 .721 1. . . . . . . .
corr x3 .676 .379 1. . . . . . .
corr x4 .149 .403 .450 1. . . . . .
corr x5 .422 .384 .445 .411 1. . . . .
corr x6 .343 .456 .243 .308 .531 1. . . .
corr x7 .115 .225 .201 .481 .373 .198 1. . .
corr x8 .213 .237 .434 .503 .267 .333 .355 1. .
corr x9 .236 .257 .159 .246 .126 .235 .601 .512 1.
mean . 21.3 22.3 17.2 23.4 22.1 15.6 18.7 20.1 19.7
std . 1.2 1.4 .87 1.33 2.2 1.4 2.3 2.1 1.8
n . 21 21 21 21 21 21 21 21 21
;
data g2(type=corr);
Input _type_ $ 18 _name_ $ 911 x1x9;
datalines;
corr x1 1. . . . . . . . .
corr x2 .733 1. . . . . . . .
corr x3 .576 .388 1. . . . . . .
corr x4 .209 .414 .425 1. . . . . .
corr x5 .412 .286 .461 .398 1. . . . .
corr x6 .323 .399 .212 .302 .522 1. . . .
corr x7 .215 .295 .188 .467 .334 .232 1. . .
corr x8 .204 .257 .462 .522 .298 .355 .372 1. .
corr x9 .245 .272 .177 .301 .156 .246 .578 .422 1.
mean . 22.1 19.8 16.9 23.3 21.9 17.3 17.9 19.1 19.8
std . 1.3 1.3 .99 1.25 2.1 1.3 2.2 2.0 1.5
n . 22 22 22 22 22 22 22 22 22
;
data g3(type=corr);
Input _type_ $ 18 _name_ $ 911 x1x9;
datalines;
corr x1 1. . . . . . . . .
corr x2 .699 1. . . . . . . .
corr x3 .488 .328 1. . . . . . .
corr x4 .235 .398 .413 1. . . . . .
corr x5 .377 .265 .471 .376 1. . . . .
corr x6 .335 .412 .265 .314 .503 1. . . .
corr x7 .243 .216 .192 .423 .369 .212 1. . .
corr x8 .217 .292 .423 .525 .219 .317 .376 1. .
corr x9 .211 .283 .152 .285 .147 .135 .633 .579 1.
mean . 22.2 20.9 15.4 25.1 22.6 16.3 19.3 20.2 19.5
std . 1.5 1.0 1.04 1.5 1.9 1.6 2.4 2.2 1.6
n . 20 20 20 20 20 20 20 20 20
;
Each of these data sets contains the information about the correlations, means, standard deviations, and sample sizes. Even
though these data sets contain correlations, by default PROC CALIS analyzes the covariances and means.
The first hypothesis to test is the equality of covariance matrices and mean vectors:
where , , and are the population covariance matrices for the three independent groups, respectively, and , , and are the population mean vectors for the three independent groups, respectively.
The following statements specify this test:
proc calis covpattern=eqcovmat meanpattern=eqmeanvec;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
In the PROC CALIS statement, the COVPATTERN=EQCOVMAT option specifies the same covariance matrix for the three groups and
the MEANPATTERN=EQMEANVEC option specifies the same mean vector for the three groups. The VAR statement specifies that x1
–9
are the variables in the hypothesis test. Next, the GROUP statements specify the data sets for the three independent groups.
You use the FITINDEX statement to limit the amount of output fit statistics to the quantities specified: the chisquare test
(CHISQ), the degrees of freedom (DF), the significance value of the test statistic (PROBCHI), the root mean square error of
approximation (RMSEA), Akaike’s information criterion (AIC), consistent Akaike’s information criterion (CAIC), and Schwarz’s
Bayesian criterion (SBC). The first three quantities are useful for the chisquare model fit test, while the rest of the fit
indices are useful for comparing competing models for the data. Because there are not many quantities in this customized fit
summary table, the NOINDEXTYPE option is used to suppress the printing of the fit index types.
Output 29.22.1 shows the general modeling information, including the sample sizes, the models for the groups, the model types, and the analysis
types.
Output 29.22.1: Modeling Information for Testing Equality of Covariance and Mean Matrices
WORK.G1 
21 
Model 1 
MSTRUCT 
Means and Covariances 
WORK.G2 
22 
Model 2 
MSTRUCT 
Means and Covariances 
WORK.G3 
20 
Model 3 
MSTRUCT 
Means and Covariances 
Output 29.22.2 shows the initial mean vector and the initial covariance matrix specifications for Model 1, which fits to Group 1. PROC CALIS
generates the mean parameter names _mean_1
, _mean_2
, …, and _mean_9
for the nine elements in the mean vector. It also generates the covariance parameter names _cov_1_1
, _cov_2_1
, …, and _cov_9_9
for the 45 nonredundant elements in the covariance matrix.
Output 29.22.2: Initial Mean Vector and Covariance Matrix for Model 1
_mean_1 
. 
_mean_2 
. 
_mean_3 
. 
_mean_4 
. 
_mean_5 
. 
_mean_6 
. 
_mean_7 
. 
_mean_8 
. 
_mean_9 
. 
Although not shown here, the initial mean vector and covariance matrices for Models 2 and 3 are exactly the same as those
shown in Output 29.22.2, as required by the equality of covariance and mean matrices in the null hypothesis .
Output 29.22.3 shows the customized fit summary table. The chisquare test statistic is 203.2605. The degrees of freedom is 108 and the
pvalue is less than 0.0001. Therefore, the hypothesis of equality in covariance and mean matrices is rejected for the three independent groups. The RMSEA index is much greater
than 0.05, which does not indicate a good model fit. Other fit indices such as AIC, CAIC, and SBC are not interpreted for
the fit of the model itself, but are useful for comparing competing models in the later discussion.
Output 29.22.3: Fit Summary for Testing : Equality of Covariance and Mean Matrices
203.2605 
108 
<.0001 
0.2100 
311.2605 
480.9897 
426.9897 
A less restrictive hypothesis is now considered. This hypothesis states the equality of covariance matrices only:
differs from in that the population means in are not constrained. To test this hypothesis, you need to change the MEANPATTERN= option to use the SATURATED keyword, as
shown in the following statements:
proc calis covpattern=eqcovmat meanpattern=saturated;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
Output 29.22.4 shows the results of the testing .
Output 29.22.4: Fit Summary for Testing : Equality of Covariance Matrices but Unconstrained Means
26.7897 
90 
1.0000 
0.0000 
170.7897 
397.0954 
325.0954 
The chisquare test statistic is 26.7897 (df = 90, p = 1.000). You cannot reject this null hypothesis about the equality of the population covariance matrices. The RMSEA value
is virtually zero, which indicates a perfect fit. Comparing the models under and , it is clear that the three groups are significantly different with regard to their mean vectors. By relaxing all the equality
constraints on the means in , is derived and is supported by the chisquare test. In addition, the RMSEA value for the model under is perfect. Because lower values of AIC, CAIC, and SBC values indicate better model fit (with the model complexity taken
into account), these indices in Output 29.22.3 and Output 29.22.4 support that the model under is better than .
However, in getting a superior model fit, might have relaxed more constraints than absolutely necessary for an optimal fit. That is, it might be possible to impose
equality constraints on only some (but not all, as in ) of the means to reach the same or even better model fit (by the RMSEA, AIC, CAIC, or SBC criterion) than the model under
. But how can you determine this set of constrained means?
To answer this question, you conduct an exploratory analysis of the data by using some model modification techniques. Models
established from exploratory analysis should be validated by external data in the future. However, this example demonstrates
the exploratory part only.
Beginning with the model under , you can manually take away some particular constraints on the means and explore whether the revised model improves the fit.
If the revised model fits better, you can repeat the process until you cannot improve more. Ultimately, you might be able
to find the “best” model between the models specified under and . Such an exploratory analysis is laborious, considering the vast possibilities of constraints on the nine variable means
in three independent groups that you could attempt to release. Fortunately, PROC CALIS provides some model modification statistics,
called the LM (Lagrange multiplier) statistics, to assist this kind of exploratory analysis.
The following statements specify the model under , but now with the MODIFICATION option added to the PROC CALIS statement:
proc calis covpattern=eqcovmat meanpattern=eqmeanvec modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
The MODIFICATION option requests the socalled LM (Lagrange multiplier) statistics for releasing the parameter constraints.
These constraints include the crossgroup or withingroup constraints and the fixed values in the model. For the model under
, the covariances and the means are all constrained across groups. These are the equality constraints that you would like
to release to obtain a better model fit. Output 29.22.5 shows the results of the LM statistics for releasing these equality constraints in variances, covariances, and means.
Output 29.22.5: Lagrange Multiplier Statistics for Releasing the Equality Constraints
0.01137 
0.9151 
0.0178 
0.0355 
1.00150 
0.3169 
0.1729 
0.3212 
1.28632 
0.2567 
0.1818 
0.3923 
2.19353 
0.1386 
0.2038 
0.4076 
0.77014 
0.3802 
0.1253 
0.2327 
0.36128 
0.5478 
0.0796 
0.1718 
3.12065 
0.0773 
0.4344 
0.8687 
0.05704 
0.8112 
0.0609 
0.1132 
4.14151 
0.0418 
0.4817 
1.0395 
0.00672 
0.9347 
0.00888 
0.0178 
2.23758 
0.1347 
0.1681 
0.3122 
2.10455 
0.1469 
0.1512 
0.3264 
2.18538 
0.1393 
0.1940 
0.3881 
3.14532 
0.0761 
0.2416 
0.4487 
0.10264 
0.7487 
0.0405 
0.0874 
1.56813 
0.2105 
0.1815 
0.3630 
0.66118 
0.4161 
0.1223 
0.2272 
4.42160 
0.0355 
0.2934 
0.6332 
0.31691 
0.5735 
0.0667 
0.1333 
0.32615 
0.5679 
0.0702 
0.1304 
0.0002277 
0.9880 
0.00172 
0.00371 
0.73377 
0.3917 
0.1242 
0.2484 
0.53196 
0.4658 
0.1097 
0.2038 
0.01445 
0.9043 
0.0168 
0.0362 
0.0000258 
0.9959 
0.000547 
0.00109 
0.24892 
0.6178 
0.0558 
0.1036 
0.25646 
0.6126 
0.0525 
0.1134 
0.04412 
0.8336 
0.0361 
0.0722 
0.52198 
0.4700 
0.1288 
0.2392 
0.90948 
0.3403 
0.1577 
0.3403 
0.0008607 
0.9766 
0.00477 
0.00953 
0.01238 
0.9114 
0.0188 
0.0348 
0.00712 
0.9328 
0.0132 
0.0285 
0.10637 
0.7443 
0.0649 
0.1297 
0.00631 
0.9367 
0.0164 
0.0304 
0.16971 
0.6804 
0.0789 
0.1702 
0.06645 
0.7966 
0.0385 
0.0771 
0.0008275 
0.9771 
0.00446 
0.00829 
0.05370 
0.8167 
0.0334 
0.0720 
0.24212 
0.6227 
0.0809 
0.1617 
0.04459 
0.8328 
0.0360 
0.0669 
0.07959 
0.7779 
0.0446 
0.0963 
0.01778 
0.8939 
0.0431 
0.0862 
0.08223 
0.7743 
0.0962 
0.1787 
0.18417 
0.6678 
0.1336 
0.2883 
0.29558 
0.5867 
0.0721 
0.1442 
0.26589 
0.6061 
0.0710 
0.1318 
1.16570 
0.2803 
0.1378 
0.2974 
0.00228 
0.9619 
0.00780 
0.0156 
1.00319 
0.3165 
0.1697 
0.3152 
0.95767 
0.3278 
0.1538 
0.3320 
1.39116 
0.2382 
0.1513 
0.3027 
0.08741 
0.7675 
0.0394 
0.0731 
0.79586 
0.3723 
0.1102 
0.2378 
0.46031 
0.4975 
0.0947 
0.1894 
0.04254 
0.8366 
0.0299 
0.0555 
0.22665 
0.6340 
0.0640 
0.1381 
0.14991 
0.6986 
0.0700 
0.1399 
0.04723 
0.8280 
0.0408 
0.0757 
0.02874 
0.8654 
0.0295 
0.0636 
0.22550 
0.6349 
0.1079 
0.2158 
0.04390 
0.8340 
0.0494 
0.0918 
0.48451 
0.4864 
0.1523 
0.3286 
0.50774 
0.4761 
0.1203 
0.2406 
0.01246 
0.9111 
0.0196 
0.0363 
0.36926 
0.5434 
0.0988 
0.2131 
0.01235 
0.9115 
0.0228 
0.0455 
0.16400 
0.6855 
0.0861 
0.1598 
0.09159 
0.7622 
0.0597 
0.1288 
0.16844 
0.6815 
0.0644 
0.1288 
0.15095 
0.6976 
0.0633 
0.1175 
0.0003079 
0.9860 
0.00265 
0.00572 
0.22542 
0.6349 
0.0776 
0.1551 
0.00754 
0.9308 
0.0147 
0.0273 
0.15376 
0.6950 
0.0617 
0.1331 
0.07831 
0.7796 
0.0631 
0.1262 
0.07552 
0.7835 
0.0643 
0.1195 
3.293E6 
0.9986 
0.000394 
0.00085 
0.13810 
0.7102 
0.0726 
0.1452 
0.0001086 
0.9917 
0.00211 
0.00392 
0.14999 
0.6985 
0.0729 
0.1572 
0.09334 
0.7600 
0.1051 
0.2101 
0.00128 
0.9714 
0.0128 
0.0237 
0.11994 
0.7291 
0.1147 
0.2474 
0.04800 
0.8266 
0.0353 
0.0706 
0.19725 
0.6569 
0.0743 
0.1379 
0.45888 
0.4981 
0.1051 
0.2268 
0.13689 
0.7114 
0.0727 
0.1453 
0.31671 
0.5736 
0.1147 
0.2130 
0.04084 
0.8398 
0.0382 
0.0825 
0.37615 
0.5397 
0.0904 
0.1808 
0.00452 
0.9464 
0.0103 
0.0191 
0.47678 
0.4899 
0.0980 
0.2114 
0.00989 
0.9208 
0.0150 
0.0300 
0.01001 
0.9203 
0.0157 
0.0291 
0.04138 
0.8388 
0.0296 
0.0638 
0.01378 
0.9066 
0.0267 
0.0533 
0.03154 
0.8590 
0.0419 
0.0778 
0.09063 
0.7634 
0.0659 
0.1421 
0.0007193 
0.9786 
0.00510 
0.0102 
0.01293 
0.9095 
0.0224 
0.0417 
0.02067 
0.8857 
0.0263 
0.0568 
0.16543 
0.6842 
0.0952 
0.1904 
0.29902 
0.5845 
0.1328 
0.2467 
0.02206 
0.8819 
0.0335 
0.0722 
0.00581 
0.9392 
0.0244 
0.0487 
0.00694 
0.9336 
0.0276 
0.0513 
0.0000660 
0.9935 
0.00250 
0.00539 
0.19272 
0.6607 
0.0532 
0.1063 
0.01910 
0.8901 
0.0174 
0.0323 
0.34408 
0.5575 
0.0684 
0.1476 
0.09017 
0.7640 
0.0446 
0.0892 
0.26496 
0.6067 
0.0794 
0.1474 
0.04994 
0.8232 
0.0320 
0.0690 
0.44236 
0.5060 
0.0758 
0.1516 
0.12761 
0.7209 
0.0422 
0.0784 
0.09470 
0.7583 
0.0338 
0.0728 
0.04619 
0.8298 
0.0260 
0.0520 
0.22996 
0.6316 
0.0602 
0.1117 
0.07502 
0.7842 
0.0319 
0.0688 
0.02807 
0.8669 
0.0279 
0.0557 
0.0006585 
0.9795 
0.00443 
0.00823 
0.02058 
0.8859 
0.0230 
0.0496 
0.03989 
0.8417 
0.0282 
0.0563 
0.15069 
0.6979 
0.0568 
0.1055 
0.36051 
0.5482 
0.0815 
0.1759 
0.03398 
0.8537 
0.0284 
0.0567 
0.05802 
0.8097 
0.0385 
0.0714 
0.00362 
0.9520 
0.00891 
0.0192 
0.06050 
0.8057 
0.0391 
0.0781 
0.56151 
0.4537 
0.1235 
0.2294 
0.26945 
0.6037 
0.0794 
0.1713 
0.13296 
0.7154 
0.0655 
0.1310 
0.00130 
0.9712 
0.00673 
0.0125 
0.16526 
0.6844 
0.0703 
0.1517 
11.09173 
0.0009 
0.3453 
0.6906 
1.21196 
0.2709 
0.1184 
0.2200 
5.04550 
0.0247 
0.2242 
0.4838 
21.46921 
<.0001 
0.5837 
1.1675 
15.27776 
<.0001 
0.5110 
0.9490 
0.47301 
0.4916 
0.0834 
0.1800 
4.41967 
0.0355 
0.2034 
0.4067 
6.37770 
0.0116 
0.2535 
0.4708 
22.27732 
<.0001 
0.4395 
0.9485 
3.26860 
0.0706 
0.1904 
0.3807 
0.03260 
0.8567 
0.0197 
0.0366 
4.06935 
0.0437 
0.2045 
0.4413 
0.22210 
0.6374 
0.0681 
0.1362 
1.50172 
0.2204 
0.1837 
0.3412 
0.60673 
0.4360 
0.1083 
0.2338 
1.61486 
0.2038 
0.1539 
0.3078 
6.72912 
0.0095 
0.3260 
0.6055 
1.88248 
0.1701 
0.1600 
0.3452 
0.14035 
0.7079 
0.0558 
0.1116 
0.11034 
0.7398 
0.0514 
0.0954 
0.00153 
0.9688 
0.00560 
0.0121 
0.12603 
0.7226 
0.0510 
0.1019 
1.96607 
0.1609 
0.2089 
0.3880 
1.16200 
0.2811 
0.1490 
0.3215 
0.05301 
0.8179 
0.0248 
0.0496 
0.97965 
0.3223 
0.1106 
0.2054 
0.61083 
0.4345 
0.0810 
0.1748 
To use the results of this table, you look for parameters that have large LM statistics (in the LM Stat column). Equivalently,
you can look for parameters that have small pvalues (in the Pr > ChiSq column). Loosely speaking, an LM statistic estimates the reduction of model fit chisquare statistic
if you release the constraint on the corresponding parameter. The pvalue indicates whether the improvement would be significant. Therefore, releasing those parameters with a high LM statistic
and small pvalue would be the key to model improvements. Bear in mind that the LM statistics are linear approximations and they might
not be very accurate as estimates of the actual model improvement, which could only be accessed when you refit the model with
the particular constraint released. Nonetheless, the LM statistics could still be very useful because they show which constraints
could potentially improve the model the most.
Output 29.22.5 shows the results from releasing the constraints on the variances and covariances first. Each constrained element of the
covariance matrix has three rows, respectively, for the three models (or groups). For example, the first parameter is _cov_1_1
, which is the same variance parameter for x1
in the three models. The first row shows that if you release the variance of x1
in Model 1 from the constraint (while keeping the variances of x1
being constrained between Models 2 and 3), the LM statistic is 0.01127, and the corresponding pvalue is 0.9155. This means that the model fit improvement would be very small and so you do not expect a significant model
fit improvement by releasing this constraint. The columns entitled “Changes” show the estimated parameter changes in the original parameters (that is, _cov_1_1
for Models 2 and 3) and in the released parameter (that is, the new parameter for the variance of x1
in Model 1) if you release the corresponding equality constraint. These two “Changes” columns are not very useful for the present purpose.
Looking through the results for the variance and covariance constraints, you can see that almost all the associated pvalues are large (that is, as compared with the conventional 0.05 level for significance). Therefore, all these constraints
on variances and covariances would not improve the model fit significantly. In contrast, the constraints on the means show
that several of them could be released for a sizable model fit improvement. The largest LM statistic in the table is the one
for _mean_3
in Model 3. The LM statistic is 22.27678 and its corresponding pvalue is less than 0.0001. This means that if the mean of x3
in Model 3 were not constrained with the means of x3
in Models 1 and 2, you would have expected a reduction in the model fit chisquare statistic that is estimated at 22.27678.
Other notable LM statistics are those for _mean_1
in Model 1, _mean_2
in Model 1 or 2, and _mean_6
in Model 2.
Two important points are noted about the use of the LM statistics. First, the LM statistics are not additive. You cannot expect
that the total reduction in model fit chisquare for releasing a particular set of parameter constraints is the sum of the
corresponding LM statistics. Second, once you release a particular constraint and refit the model, the LM statistics in the
revised model might not follow the same pattern as those LM statistics in the original model. Basically, these are due to
the nonlinearity of the fit function and the dependence of the parameter estimates. Therefore, in order to find the best model
for the data, it would be more sensible to adopt a oneatatime approach to release the constraints. That is, you release
one constraint at a time and refit the model to see if you can release more constraints to improve the model fit.
According to the results of LM statistics in Output 29.22.5, you first release the constraint on the _mean_3
parameter, which is for the mean of x3
in Model 3. The following statements fit such a model:
proc calis modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
model 1 / group = 1;
mstruct;
matrix _cov_ = cov01cov45;
matrix _mean_ = mean1mean9;
model 2 / group = 2;
refmodel 1;
model 3 / group = 3;
refmodel 1;
renameparm mean3=mean3_mdl3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
Because the revised model is no longer a supported builtin MSTRUCT model, you cannot use the MEANPATTERN= or the COVPATTERN=
options any more. Instead, you now use the MSTRUCT modeling language to specify the covariance and mean patterns. Model 1,
which fits to Group 1, is an MSTRUCT model with variance and covariance parameters cov01
–cov45
and mean parameters mean1
–mean9
. Model 2, which fits to Group 2, refers to the specifications of Model 1, as indicated in a REFMODEL statement. Hence, Model
1 and Model 2 are completely constrained in variances, covariances, and means. Model 3, which fits to Group 3, also refers
to the specifications of Model 1, as indicated in another REFMODEL statement. However, the RENAMEPARM statement renames the
parameter mean3
in the reference model (that is, Model 1) to a new name mean3_mdl3
. As a results, all variance, covariance, and mean parameters except one in Model 3 are constrained to be the same as those
in Model 1. The mean of x3
in Model 3 is the only parameter that is not constrained with any other parameters. This forms the first revised model from
. The MODIFICATION option is specified again to determine whether a further model fit improvement is possible.
Output 29.22.6 shows the modeling information of the first revised model. It shows that Models 2 and 3 make references to Model 1. Therefore,
parameters between models are constrained by referencing.
Output 29.22.6: Modeling Information for The First Revised Model
WORK.G1 
21 
Model 1 
MSTRUCT 

Means and Covariances 
WORK.G2 
22 
Model 2 
MSTRUCT 
Model 1 
Means and Covariances 
WORK.G3 
20 
Model 3 
MSTRUCT 
Model 1 
Means and Covariances 
Output 29.22.7 shows the initial specifications of the means, variances, and covariances in Model 1.
Output 29.22.7: Initial Mean Vector and Covariance Matrix for Model 1 in the First Revised Model
mean1 
. 
mean2 
. 
mean3 
. 
mean4 
. 
mean5 
. 
mean6 
. 
mean7 
. 
mean8 
. 
mean9 
. 
Output 29.22.8 shows the initial specifications of the means in Model 2. The mean parameters in Model 2 are exactly the same as those in
Model 1, as shown in Output 29.22.7. The variance and covariance parameters in Model 2 are also exactly the same as those in Model 1, but are not shown here
to conserve space.
Output 29.22.8: Initial Mean Vector for Model 2 in the First Revised Model
mean1 
. 
mean2 
. 
mean3 
. 
mean4 
. 
mean5 
. 
mean6 
. 
mean7 
. 
mean8 
. 
mean9 
. 
Output 29.22.9 shows the initial specifications of the means in Model 3. All but one mean parameter in Model 3 are exactly the same as those
in Models 1 and 2, as shown in Output 29.22.7 and Output 29.22.8, respectively. The mean for x3
in Model 3 is mean3_mdl3
, which is now a distinct parameter, and therefore it is not constrained with any other parameters in the first or the second
models for Groups 1 or 2. However, the variance and covariance parameters in Model 3 are exactly the same as those in Model
1. They are not shown here to conserve space.
Output 29.22.9: Initial Mean Vector for Model 3 in the First Revised Model
mean1 
. 
mean2 
. 
mean3_mdl3 
. 
mean4 
. 
mean5 
. 
mean6 
. 
mean7 
. 
mean8 
. 
mean9 
. 
Output 29.22.10 shows the fit summary of the first revised model. The model fit chisquare is 148.8865, which drops quite a bit from the
original model under . The pvalue of the model fit chisquare is 0.0046, which is statistically significant. The RMSEA value is 0.1399, which is also
a sizable improvement. All the AIC, CAIC, and SBC values are reduced, indicating better model fit than the model under .
Output 29.22.10: Fit Summary for the First Revised Model
148.8865 
107 
0.0046 
0.1399 
258.8865 
431.7589 
376.7589 
Output 29.22.11 shows the LM statistics for releasing the equality constraints in the first revised model. Almost all of the results for
the variance and covariance constraints are omitted because their LM statistics are not significant. However, Output 29.22.11 shows all the LM statistics for releasing the constraints in means. The mean of x2
in Model 2 has the largest LM statistic at 26.25044.
Output 29.22.11: LM Statistics for Releasing the Equality Constraints in the First Revised Model
0.64999 
0.4201 
0.1050 
0.2100 
0.41758 
0.5181 
0.0874 
0.1622 
2.18923 
0.1390 
0.1855 
0.4004 
. 



. 



. 



9.26674 
0.0023 
0.2872 
0.5745 
3.00599 
0.0830 
0.1702 
0.3160 
2.13787 
0.1437 
0.1481 
0.3196 
26.25115 
<.0001 
0.6568 
1.3135 
12.34638 
0.0004 
0.4674 
0.8680 
2.52683 
0.1119 
0.1962 
0.4234 
0.58891 
0.4428 
0.0787 
0.0827 
0.58891 
0.4428 
0.0827 
0.0787 
6.59009 
0.0103 
0.2746 
0.5493 
0.51343 
0.4737 
0.0796 
0.1478 
11.61610 
0.0007 
0.3586 
0.7739 
0.52967 
0.4667 
0.1042 
0.2084 
0.22294 
0.6368 
0.0702 
0.1304 
0.06889 
0.7930 
0.0374 
0.0807 
1.16656 
0.2801 
0.1270 
0.2540 
5.29599 
0.0214 
0.2810 
0.5218 
1.69412 
0.1931 
0.1518 
0.3275 
0.03791 
0.8456 
0.0291 
0.0582 
0.44510 
0.5047 
0.1036 
0.1923 
0.23804 
0.6256 
0.0704 
0.1520 
0.39420 
0.5301 
0.0883 
0.1765 
0.24231 
0.6225 
0.0719 
0.1335 
0.01951 
0.8889 
0.0200 
0.0431 
0.00156 
0.9685 
0.00423 
0.00846 
1.06866 
0.3012 
0.1150 
0.2136 
1.05210 
0.3050 
0.1065 
0.2297 
You now modify the preceding statements to specify the second revised model, as shown in the following statements:
proc calis modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
model 1 / group = 1;
mstruct;
matrix _cov_ = cov01cov45;
matrix _mean_ = mean1mean9;
model 2 / group = 2;
refmodel 1;
renameparm mean2=mean2_new; /* constraint a */
model 3 / group = 3;
refmodel 1;
renameparm mean2=mean2_new, /* constraint a */
mean3=mean3_mdl3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
This second revised model must not constrain the mean of x2
in Model 1 with any parameters. A straightforward way to do this is to rename the mean2
parameter to a unique name in Model 1. However, for the current specification it is more convenient to rename the mean2
parameter in Models 2 and 3 to another name. In the specification of the second revised model, Models 2 and 3 still make
references to Model 1. However, in the respective RENAMEPARM statements, both Model 2 and 3 rename the mean2
parameter that is referenced from Model 1 to the new name mean2_new
. This way the mean for x2
in Model 1 is not constrained with the means of x2
in Models 2 and 3. But the means for x2
in Models 2 and 3 are still constrained to be equal by the same parameter mean2_new
. Output 29.22.12 shows the fit summary of the second revised model.
Output 29.22.12: Fit Summary for the Second Revised Model
86.3927 
106 
0.9183 
0.0000 
198.3927 
374.4083 
318.4083 
Again, a sizable improvement over the first revised model is shown in the second revised model. The model fit chisquare statistic
is no longer significant (p = 0.9183), and the RMSEA value is perfect at 0. Large drops in the AIC, CAIC, and SBC values are also observed.
Output 29.22.13 suggests that the mean of x6
in Model 2 (which has the largest LM statistic at 11.41243) could be released from the equality constraints to achieve the
largest model improvement over the current model.
Output 29.22.13: LM Statistics for Releasing the Equality Constraints in the Second Revised Model
2.77024 
0.0960 
0.1384 
0.2770 
0.28728 
0.5920 
0.0462 
0.0860 
5.00087 
0.0253 
0.1791 
0.3864 
. 



. 



. 



2.75437 
0.0970 
0.1646 
0.3292 
3.21093 
0.0731 
0.1511 
0.2806 
0.24923 
0.6176 
0.0424 
0.0915 
0.74338 
0.3886 
0.0877 
0.0934 
0.74338 
0.3886 
0.0934 
0.0877 
6.17449 
0.0130 
0.2672 
0.5343 
0.02087 
0.8851 
0.0146 
0.0272 
4.71344 
0.0299 
0.2072 
0.4470 
1.65517 
0.1983 
0.1853 
0.3706 
1.16118 
0.2812 
0.1606 
0.2982 
0.04040 
0.8407 
0.0287 
0.0618 
5.03834 
0.0248 
0.2712 
0.5423 
11.41247 
0.0007 
0.4217 
0.7831 
1.51175 
0.2189 
0.1460 
0.3150 
0.32382 
0.5693 
0.0853 
0.1706 
0.82184 
0.3646 
0.1410 
0.2619 
0.12513 
0.7235 
0.0512 
0.1104 
2.39207 
0.1220 
0.2210 
0.4420 
1.58292 
0.2083 
0.1867 
0.3467 
0.08641 
0.7688 
0.0427 
0.0922 
0.00682 
0.9342 
0.00886 
0.0177 
1.20949 
0.2714 
0.1225 
0.2274 
1.10016 
0.2942 
0.1089 
0.2349 
4.47814 
0.0343 
0.2983 
0.2661 
4.47814 
0.0343 
0.2661 
0.2983 
The process of model refitting should now become familiar. You modify the previous model to release the constraint on the
mean of x6
in Model 2. As a result, the third revised model is specified by the following statements:
proc calis modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
model 1 / group = 1;
mstruct;
matrix _cov_ = cov01cov45;
matrix _mean_ = mean1mean9;
model 2 / group = 2;
refmodel 1;
renameparm mean2=mean2_new, /* constraint a */
mean6=mean6_mdl2;
model 3 / group = 3;
refmodel 1;
renameparm mean2=mean2_new, /* constraint a */
mean3=mean3_mdl3;
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
The only modification from the previous specification is to rename mean6
to mean6_mdl2
in the RENAMEPARM statement of Model 2. Output 29.22.14 shows the model fit summary of the third revised model.
Output 29.22.14: Fit Summary for the Third Revised Model
68.7869 
105 
0.9976 
0.0000 
182.7869 
361.9456 
304.9456 
The model improvement over the second revised model is still notable in the third revised model. The chisquare value drops
about 20 points in the third revised model. The AIC, CAIC, and the SBC values are reduced notably, though not as impressively
as with the previous improvements.
Output 29.22.15 suggests that the mean of x4
in Model 1 (which has the largest LM statistic at 7.01946) could be released from the equality constraint to improve model
fit further.
Output 29.22.15: LM Statistics for Releasing the Equality Constraints in the Third Revised Model
2.43374 
0.1187 
0.1342 
0.2684 
0.19037 
0.6626 
0.0390 
0.0723 
4.11402 
0.0425 
0.1679 
0.3625 
. 



. 



. 



6.15722 
0.0131 
0.2550 
0.5101 
6.05791 
0.0138 
0.2109 
0.3917 
0.29302 
0.5883 
0.0463 
0.0999 
2.89780 
0.0887 
0.1796 
0.1889 
2.89780 
0.0887 
0.1889 
0.1796 
7.01915 
0.0081 
0.2850 
0.5701 
0.04916 
0.8245 
0.0226 
0.0419 
5.05102 
0.0246 
0.2148 
0.4635 
0.21231 
0.6450 
0.0672 
0.1345 
0.07502 
0.7842 
0.0443 
0.0822 
0.55031 
0.4582 
0.1059 
0.2285 
0.07013 
0.7911 
0.0503 
0.0486 
0.07013 
0.7911 
0.0486 
0.0503 
0.98902 
0.3200 
0.1513 
0.3025 
2.42355 
0.1195 
0.2463 
0.4575 
0.34231 
0.5585 
0.0858 
0.1850 
1.58481 
0.2081 
0.1786 
0.3572 
0.81634 
0.3663 
0.1347 
0.2502 
0.14503 
0.7033 
0.0549 
0.1184 
0.13504 
0.7133 
0.0399 
0.0797 
2.54369 
0.1107 
0.1796 
0.3335 
1.61681 
0.2035 
0.1337 
0.2885 
3.21203 
0.0731 
0.2484 
0.2280 
3.21203 
0.0731 
0.2280 
0.2484 
To make the mean parameter for x4
in Model 1 unique, the mean parameters for x4
in Models 2 and 3 are renamed from mean4
to mean4_new
, as shown in the following statements:
proc calis modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
model 1 / group = 1;
mstruct;
matrix _cov_ = cov01cov45;
matrix _mean_ = mean1mean9;
model 2 / group = 2;
refmodel 1;
renameparm mean2=mean2_new, /* constraint a */
mean4=mean4_new, /* constraint b */
mean6=mean6_mdl2;
model 3 / group = 3;
refmodel 1;
renameparm mean2=mean2_new, /* constraint a */
mean3=mean3_mdl3,
mean4=mean4_new; /* constraint b */
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
This forms the fourth revised model. Output 29.22.16 shows the fit summary of this revised model. Again, the chisquare, AIC, CAIC, and SBC values all show improvements, as compared
with the third revised model. However, the improvements do seem to slow down. For example, the CAIC value drops from 361.95
to the current value at 358.43—a mere 3 points reduction. The SBC value drops from 304.95 to the current value at 300.43—a
mere 4 points reduction. These small reductions indicate that you might soon reach the point that no more model fit improvement
would be possible with additional release of parameter constraints.
Output 29.22.16: Fit Summary for the Fourth Revised Model
60.1265 
104 
0.9998 
0.0000 
176.1265 
358.4283 
300.4283 
Output 29.22.17 suggests that the mean of x1
in Model 1 (which has the largest LM statistic at 6.45785) could be released from the equality constraint to achieve the
largest model improvement over the current model.
Output 29.22.17: LM Statistics for Releasing the Equality Constraints in the Fourth Revised Model
2.60531 
0.1065 
0.1376 
0.2751 
0.28122 
0.5959 
0.0469 
0.0871 
4.75001 
0.0293 
0.1788 
0.3859 
. 



. 



. 



6.45761 
0.0110 
0.2616 
0.5232 
5.00991 
0.0252 
0.1921 
0.3568 
0.05931 
0.8076 
0.0209 
0.0452 
1.53300 
0.2157 
0.1298 
0.1406 
1.53300 
0.2157 
0.1406 
0.1298 
0.09749 
0.7549 
0.0457 
0.0913 
0.19688 
0.6572 
0.0716 
0.1330 
0.56583 
0.4519 
0.1071 
0.2310 
0.35800 
0.5496 
0.1141 
0.1113 
0.35800 
0.5496 
0.1113 
0.1141 
4.53367E6 
0.9983 
0.000350 
0.00070 
0.96363 
0.3263 
0.1572 
0.2920 
1.00890 
0.3152 
0.1486 
0.3208 
0.20289 
0.6524 
0.0676 
0.1351 
0.12445 
0.7243 
0.0525 
0.0974 
0.00590 
0.9388 
0.0110 
0.0237 
0.05893 
0.8082 
0.0271 
0.0542 
1.63723 
0.2007 
0.1448 
0.2689 
2.44241 
0.1181 
0.1652 
0.3565 
3.05068 
0.0807 
0.2396 
0.2246 
3.05068 
0.0807 
0.2246 
0.2396 
1.81983 
0.1773 
0.2306 
0.2003 
1.81983 
0.1773 
0.2003 
0.2306 
To make the mean parameter for x1
in Model 1 unique, the mean parameters for x1
in Models 2 and 3 are renamed from mean1
to mean1_new
, as shown in the following statements:
proc calis modification;
var x1x9;
group 1 / data=g1;
group 2 / data=g2;
group 3 / data=g3;
model 1 / group = 1;
mstruct;
matrix _cov_ = cov01cov45;
matrix _mean_ = mean1mean9;
model 2 / group = 2;
refmodel 1;
renameparm mean1=mean1_new, /* constraint c */
mean2=mean2_new, /* constraint a */
mean4=mean4_new, /* constraint b */
mean6=mean6_mdl2;
model 3 / group = 3;
refmodel 1;
renameparm mean1=mean1_new, /* constraint c */
mean2=mean2_new, /* constraint a */
mean3=mean3_mdl3,
mean4=mean4_new; /* constraint b */
fitindex NoIndexType On(only)=[chisq df probchi rmsea aic caic sbc];
run;
This forms the fifth revised model. Output 29.22.18 shows the fit summary of the fifth revised model. Again, the chisquare, AIC, CAIC, and SBC values all show improvements,
as compared with the fourth revised model. However, the improvements slow down even more. For example, the CAIC value drops
from 358.43 to the current value at 356.32. The SBC value drops from 300.43 to the current value at 297.32. Because the model
fit does not improve much, this is the point where you would cease to release more equality constraints for improving the
model fit.
Output 29.22.18: Fit Summary for the Fifth Revised Model
52.8821 
103 
1.0000 
0.0000 
170.8821 
356.3270 
297.3270 
Output 29.22.19 does not suggest the release of any equality constraints on the means, because all the pvalues for the LM statistics are not significant (that is, all are greater than 0.05). Therefore, the same suggestion from
examining the model fit improvements of the fifth revised model echoes here: this is the point that the “best” model for the data is found.
Output 29.22.19: LM Statistics for Releasing the Equality Constraints in the Fifth Revised Model
4.06279 
0.0438 
0.1590 
0.3180 
0.48735 
0.4851 
0.0571 
0.1061 
7.60892 
0.0058 
0.2095 
0.4520 
. 



. 



. 



0.08363 
0.7724 
0.0312 
0.0382 
0.08363 
0.7724 
0.0382 
0.0312 
0.02394 
0.8770 
0.0229 
0.0458 
0.47076 
0.4926 
0.1113 
0.2067 
0.26015 
0.6100 
0.0728 
0.1571 
0.97521 
0.3234 
0.1893 
0.1892 
0.97521 
0.3234 
0.1892 
0.1893 
0.03746 
0.8465 
0.0319 
0.0638 
1.10428 
0.2933 
0.1683 
0.3126 
0.79474 
0.3727 
0.1321 
0.2851 
0.86792 
0.3515 
0.1426 
0.2852 
0.47493 
0.4907 
0.1038 
0.1928 
0.03722 
0.8470 
0.0276 
0.0595 
0.12190 
0.7270 
0.0401 
0.0801 
2.66768 
0.1024 
0.1869 
0.3472 
1.78114 
0.1820 
0.1417 
0.3058 
1.28034 
0.2578 
0.1794 
0.1359 
1.28034 
0.2578 
0.1359 
0.1794 
2.53131 
0.1116 
0.2117 
0.2112 
2.53131 
0.1116 
0.2112 
0.2117 
2.25832 
0.1329 
0.2558 
0.2253 
2.25832 
0.1329 
0.2253 
0.2558 
To see where the fifth revised model (equality in the covariance matrix and partial equality in the means) stands between
the models under (equality in the covariance and mean matrices) and (equality in the covariance matrix only), the following table shows the fit statistics of these three models:


“Fifth”


Chisquare

203.2605

52.8821

26.7897

Chisquare DF

108

103

90

Pr > chisquare

<0.0001

1.0000

1.0000

RMSEA estimate

0.2100

0.0000

0.0000

Akaike information criterion

311.2605

170.8821

170.7897

Bozdogan CAIC

480.9898

356.3270

397.0954

Schwarz Bayesian criterion

426.9898

297.3270

325.0954

The fifth revised model is labeled “Fifth” in the table. Compared with the model under , the fifth revised model is clearly superior. It uses only five more parameters (or five fewer degrees of freedom), but the
improvement in the model fit chisquare and the RMSEA value are huge. The AIC, CAIC, and SBC are also much better.
Compared with the model under , the fifth revised model appears to be inferior in only the chisquare model fit statistic, although both models already
have the highest possible pvalue at 1.000 and smallest possible RMSEA value at 0. However, the model under uses 13 more parameters (or it has 13 fewer degrees of freedom), and hence it is more complex. In fact, because the model
fit chisquare value does not take model complexity into account, it is often criticized as the basis for choosing competing
models for the data. In contrast, the AIC, CAIC, and SBC measures take model complexity into account, and they are more reasonable
as the basis for choosing competing models. Although the AIC values for the fifth revised model and the model under are very close, the CAIC and SBC values clearly favor the fifth revised model. Therefore, according to the CAIC and SBC criteria,
the fifth revised model, which is a model with partial equality constraints on the means, is actually better than the model
with all the means being unconstrained (that is, under ) for the current data with three independent groups.