LISREL Examples Guide

Table of Contents
1 Using Existing SIMPLIS Syntax Files ............................................................................ 3

1.1 Changing the Appearance of the Output .............................................................. 8
2 Using Existing LISREL Syntax Files .............................................................................. 8
2.1 Creating Syntax by Drawing Path Diagrams ..................................................... 12
2.2 Title, Labels and Data ........................................................................................ 14
2.3 Drawing the Path Diagram ................................................................................. 18
3 Multi-Sample Analyses Using Path Diagrams .............................................................. 24
3.1 Testing Equality of Factor Structures................................................................. 25
3.2 Drawing the Path Diagram ................................................................................. 31
4 Analysis of Ordinal Data using an asymptotic covariance matrix. ................................ 36
4.1 Calculation of Correlations and Asymptotic Covariance Matrix ....................... 36
4.2 Drawing a Path Diagram .................................................................................... 38
4.3 Reading Variable Names from the DSF File ..................................................... 39
4.4 Creating LISREL and/or SIMPLIS Syntax from the Path Diagram .................. 44
4.5 The Basic Model and its Components ............................................................... 48
4.6 Estimation Types and the Conceptual Path Diagram ......................................... 51
4.7 Saving the Path Diagram in Graphics Format .................................................... 51
5 Empty Path Diagrams .................................................................................................... 52
5.1 Motivation .......................................................................................................... 52
5.2 Examples ............................................................................................................ 52
6 Using SIMPLIS Project Files......................................................................................... 54
6.2 Building the SIMPLIS Syntax............................................................................ 56
6.3 Building and Completing the SIMPLIS Syntax ................................................. 59
7 Using LISREL Project Files .......................................................................................... 61
7.1 Building the LISREL Syntax ............................................................................. 63
8 Analysis with Missing Data ........................................................................................... 73
8.1 Single-Group Analysis with Missing Data ........................................................ 73
8.2 Multiple-groups with Missing Data ................................................................... 80
9 Latent Variable Scores ................................................................................................... 90
9.1 Calculation of Latent Variable Scores ............................................................... 90
9.2 An Illustrative Example ..................................................................................... 91
10 Multilevel Confirmatory Factor Analysis .................................................................... 94
10.1 Import of External Data .................................................................................. 94
LISREL Examples Guide 1
10.2 Draw the Path Diagram .................................................................................. 95
10.3 Build SIMPLIS/LISREL Syntax .................................................................... 98
10.4 Fit the Model to the Data ................................................................................ 99

1 Using Existing SIMPLIS Syntax Files
The SIMPLIS examples are contained in the splex folder of LISREL see table below.
LIST OF SIMPLIS EXAMPLES IN THE \SPLEX FOLDER
Example 1: Regression of GNP Example 2: Prediction of Grade Averages

Data File: = None Data File: None
Input Files: EX1A.SPL, EX1B.SPL Input Files: EX2A.SPL, EX2B.SPL
Example 3: Union Sentiment of Textile Example 4: Ability and Aspiration
Workers Data File: EX4.COR
Data File: None Input Files: EX4A.SPL, EX4B.SPL
Input Files: EX3A.SPL, EX3B.SPL
Example 5: Nine Psychological Variables Example 6: Stability of Alienation
Data File: EX5.COR Data File: None
Input Files: EX5A.SPL, EX5B.SPL Input Files: EX6A.SPL, EX6B.SPL, EX6C.SPL
Example 7: Performance and Satisfaction Example 8: Peer Influences on Ambition
Data File: EX7.DAT Data Files: EX8.LAB, EX8.COR, EX8.STD
Input Files: EX7A.SPL, EX7B.SPL Input Files: EX8A.LS7, EX8B.LS7,
EX8C.LS7,EX8D.LS7
Example 9: Panel Model for Political Example 10: Testing Equality of Factor
Efficacy Structures
Data Files: PANEL.LAB, PANELUSA.PME, Data File: EX10.COV
PANELUSA.ACP Input Files: EX10A.SPL, EX10B.SPL,
Input Files: EX9A.SPL, EX9B.SPL EX10C.SPL, EX10D.SPL
Example 11: Parental Socioeconomic Example 12: Testing Equality of
Characteristics Regressions
Data File: None Data File: EX12.DAT
Input Files: EX11A.SPL, EX11B.SPL Input Files: EX12A.SPL, EX12B.SPL,
EX12C.SPL
Example 13: Mean Difference in Verbal Example 14: Nine Psychological Variables
Ability with Factor Means
Data File: EX12.DAT Data Files: EX14.LAB, EX14.DAT
(Note: Same Data File as for Example 12) Input File: EX14.SPL
Example 15: Regression of Verbal7 on Example 16: Head Start Summer Program
Verbal5 Data File: EX16.DAT
Data File: EX12.DAT Input Files: EX16A.SPL, EX16B.SPL,
(Note: Same Data File as for Example 12) EX16C.SPL, EX16D.SPL
Example 17: Hypothetical Model
Data Files: EX17.COV, EX17.COR
To run a SIMPLIS example, select the Open option from the File menu to obtain the Open
dialog box. Select (for example) ex10a.spl from the splex folder. Click Open when done.

Click the Run LISREL icon button to run ex10a.spl. Alternatively, use the File, Run
LISREL option.
Note that a path diagram is obtained by entering the command "Path Diagram" in the file
ex10a.spl, just below the line Sample Size = 900 as shown above. The path diagram is
displayed in the path diagram window.

A text editor window is also opened. A selection of the output contained in ex10a.out is
displayed below.
Once a path diagram is produced, the LISREL menu bar will display a different selection
of options, as shown next.

After a path diagram has been obtained, one may select the Draw menu to add paths,
change text, etc. The items on this menu correspond to the Draw toolbox displayed next to
it on the window shown below.
The default values shown on the path diagram are the parameter estimates. By selecting
the Estimations option from the View menu, one may alternatively select the t-values,
modification indices, etc.
One may change the number of decimals displayed in the path diagram by selecting
SIMPLIS outputs from the Output menu. On the dialog box displayed, change the number
of decimals from the default value of 2.
The SIMPLIS Outputs dialog box is shown below:

The LISREL syntax that corresponds to the SIMPLIS command language in ex10a.spl may
be built from the path diagram. To do so, select the Setup, Build LISREL Syntax option as
shown in the next image.
A LISREL project file (ex10a.lpj) is produced, containing the corresponding LISREL

syntax, as shown below.

1.1 Changing the Appearance of the Output
While the output file is opened, one may use the Edit menu to perform various operations
such as finding a keyword (Find), replacing text (Replace), inserting files (Insert File), etc.
When output is displayed, one may also use the Options menu to change the font type and
size of the text.
2 Using Existing LISREL Syntax Files

LISREL syntax examples are contained in the lisex folder, see tables below.

LIST OF LISREL EXAMPLES IN THE \LISEX FOLDER
Example 1: Hypothetical model estimated Example 3.1: Analysis of reader reliability

by ML in essay scoring
Data File: EX1.COV Data File: None
Input File: EX1.LIS Input Files: EX31A.LIS, EX31B.LIS,
EX31C.LIS
Example 3.2: Ability and aspiration Example 3.3: Estimating the dis-attenuated
Data File: None correlation
Input Files: EX32A.LIS, EX32B.LIS Data File: EX33.COV
Input File: EX33.LIS
Example 3.4: Nine Psychological Variables Example 4.1: Regression of GNP on
Data File: EX34.COR economic factors
Input Files: EX34A.LIS, EX34B.LIS Data File: EX41.RAW
Example 4.2: Stepwise regression Example 4.3: ANOVA and ANCOVA
Data File: EX42.COV Data File: EX43.RAW
EX43C.LIS
Example 4.4: Bivariate regression Example 4.5: Ambition and attainment
Data File: EX44.RAW Data File: None
Input File: EX44.LIS Input Files: EX45A.LIS, EX45B.LIS
Example 4.6: Klein's model I of US Example 5.1: Verbal ability in grades 4 and
economy 5
Data File: EX46.DAT Data File: EX51.COV
Input Files: EX46A.LIS, EX46B.LIS, Input File: EX51.LIS
EX46C.LIS
Example 5.2: Role behavior of farm Example 5.3: Educational attainment
managers Data File: EX53.COR
Data Files: EX52A.COV, EX52B.COV Input Files: EX53A.LIS, EX53B.LIS,
Input Files: EX52A.LIS, EX52B.LIS EX53C.LIS
Example 5.4: Social status and social Example 5.5: Peer influences on ambition
participation Data Files: EX55.LAB, EX55.COR
Data File: EX54.COR Input Files: EX55A.LIS, EX55B.LIS
Example 5.6: Performance and satisfaction Example 6.1: Three subtests of SAT
Data File: EX56.DAT Data File: None
Input File: EX56.LIS Input File: EX61.LIS
Example 6.2: Second-order factor analysis Example 6.3: The rod and frame test
Example 6.4: Stability of alienation Example 6.5: Change in verbal and
Data File: EX64.COV quantitative ability between grades 7 and 9
Input Files: EX64A.LIS, EX64B.LIS, Data File: EX65.DAT
EX64C.LIS, EX64D.LIS Input Files: EX65A.LIS, EX65B.LIS

LIST OF LISREL EXAMPLES IN THE \LISEX FOLDER (Continued)
Example 6.6: A simplex model for academic Example 7.1: Attitudes of morality and
performance equality
Data File: EX66.COR Data Files: EX71.RAW, EX71.PML,
Input Files: EX66A.LIS, EX66B.LIS, EX71.ACP
EX66C.LIS, EX66D.LIS Input Files: EX71.PRL, EX71.LIS
Example 7.2: A panel model for political Example 7.3: Factor analysis of
efficacy dichotomous variables
Data Files: PANEL.LAB, PANELUSA.RAW, Data Files: LSAT6.DAT, EX73.PML,
PANELUSA.PME, PANELUSA.ACE, EX73.ACP
PANELUSA.ACP Input Files: EX73.PRL, EX73.LIS
Input Files: EX72A.PRL, EX72B.PRL,
EX72A.LIS, EX72B.LIS
Example 7.4: Analysis of covariance Example 7.5: Estimating and testing a
matrices with WLS correlation structure
Data Files: EX74.RAW, EX74.CML, Data Files: EX75.KML, EX75.ACK
EX74.ACC Input File: EX75.LIS
Input Files: EX74.PRL, EX74.LIS
Example 8.1: Gosta's bad sample Example 8.2: Hypothesis testing and
Data Files: EX81.SIG, EX81.COV power calculation
Input Files: EX81A.LIS, EX81B.LIS, Data Files: EX82.COV, EX82.SIG
EX81C.LIS, EX81D.LIS, EX81E.LIS Input Files: EX82A.LIS, EX82B.LIS,
EX82C.LIS, EX82D.LIS
Example 9.1: Testing equality of factor Example 9.2: Testing equality of factor
structures correlation matrices
Input Files: EX91A.LIS, EX91B.LIS, Input File: EX92.LIS
EX91C.LIS, EX91D.LIS, EX91E.LIS
Example 9.3: Son's and parents' reports of Example 9.4: Subjective and objective
parental socioeconomic characteristics social class
Data File: None Data File: EX94.DAT
Example 10.1: Nine psychological Example 10.2: Head start summer program
variables with factor means Data File: EX102.DAT
Data File: EX101.DAT, EX101.LAB Input File: EX102.LIS
Example 10.3: Estimating a correlation Example 10.4: Berkeley guidance study
from incomplete data Data Files: GIRLS.COV, BOYS.COV,
Data File: EX103.DAT GIRLS.MEA, BOYS.MEA
EX104C.LIS
Additional examples are contained in the ls9ex, obsresex, missingex, nsfex, and tutorial
folders. Select the Open option from the File menu to obtain the Open dialog box. Select
(for example) the file ex31a.lis. Click Open when done. The contents of the file ex31a.lis
are displayed.

To run LISREL, click the Run LISREL icon button. Make sure that the Run LISREL icon
button is selected when running LISREL or SIMPLIS syntax and that the Run PRELIS icon
button is selected when running PRELIS syntax.
To produce the path diagram shown below, the keyword PD (Path Diagram) was entered
between the lines containing the words Esayabil and OU in ex31a.lis.

The file ex31a.out is also produced. This file may be viewed by either closing the path
diagram window or by clicking on the output window, which is partially hidden behind
the path diagram window.
2.1 Creating Syntax by Drawing Path Diagrams
LISREL allows the user to create LISREL or SIMPLIS syntax by drawing the appropriate
path diagram on the screen and then building the corresponding syntax directly from the
path diagram. This procedure will be demonstrated by fitting a CFA (Confirmatory Factor
Analysis) model to the data. To obtain correct standard errors of the estimators of the
parameters, the CFA model should be fitted to the observed covariance matrix. However,
in order to illustrate the various features of the program, the model will be fitted to the
observed correlation matrix.
In confirmatory factor analysis, one builds a model assumed to describe or account for
the empirical data in terms of relatively few parameters. The model is based on a priori
information about the data structure in the form of a specified theory or hypothesis or
knowledge from previous studies based on extensive data.
Holzinger & Swineford (1939) collected data on twenty-six psychological tests

administered to 145 seventh- and eighth-grade children in the Grant-White school in
Chicago. Six of these tests were selected and for this example it was hypothesized that
these measure two common factors: visual perception and verbal ability such that the first
three variables measure visual perception and the last three measure verbal ability. The
six selected variables and their intercorrelations are given below:
VIS PERC 1.000
CUBES 0.318 1.000
LOZENGES 0.436 0.419 1.000
PAR COMP 0.335 0.234 0.323 1.000
SEN COMP 0.304 0.157 0.283 0.722 1.000
WORDMEAN 0.326 0.195 0.350 0.714 0.685 1.000
The path diagram for this model is given below:
The corresponding SIMPLIS syntax file is:
Six Psychological Variables-A Confirmatory Factor Analysis

Observed variables
‘VIS PERC’ CUBES LOZENGES ‘PAR COMP’ ‘SEN COMP’ WORDMEAN
Correlation Matrix From File EX5.COR
Sample Size: 145
Latent Variables: Visual Verbal
Relationships:
‘VIS PERC’ - LOZENGES = Visual
‘PAR COMP’ - WORDMEAN = Verbal
Number of decimals = 4
Print Residuals
End of Problem
Some important aspects contained in the SIMPLIS syntax file given above are
o The correlation matrix is read from an external file
o There are two latent variables in the model: Visual and Verbal

o The statement Number of decimals = 4 specifies that we wish to have the results
in the output file given to an accuracy of four decimals. LISREL uses two
decimals by default.
2.2 Title, Labels and Data
Select the File, New option and click on Path Diagram. In the Save As dialog box, select a
filename and a folder in which the path diagram should be saved. For the present example
the folder chosen is splex and the file name selected is cfa6.pth. When done, click Save.
From the Output menu select SIMPLIS Outputs. The dialog box below will appear.
Customize this dialog box according to your preferences. For example, change the
Number of Decimals in the Output option to 4. Make sure that the Invoke Path Diagram
check box (the default) is checked. When done, click OK.
Before drawing the path diagram, select Toolbars from the View menu and ensure that the
items shown below are selected. These are: Toolbar, Status Bar, Typebar, Variables,
Drawing Bar. Also check the Grid Lines option.
Select the Title and Comments option from the Setup menu to obtain the Title and
Comments dialog box.
The first item on the Title and Comments dialog box is the Title for the particular analysis.
Provision is also made for any additional Comments that the user may wish to enter.
After typing in the title and (optionally) the comments, click on Next to go to the Group
Names dialog box.

Since the present example is based on the analysis of one correlation matrix, nothing is
entered in the space allowed for Group Labels and Next is clicked to go to the Labels
dialog box. The default number of variables shown on the Labels dialog box is 3, these
being CONST, VAR 1 and VAR 2. Click in the CONST textbox and change the name to VIS
PERC. Enter the label CUBES in the VAR 1 textbox and LOZENGES in the VAR 2 text box.
Move the mouse cursor into the LOZENGES textbox and use the down arrow to create a
fourth text box. Type in PAR COMP and repeat the procedure for SEN COMP and
WORMEAN.
In a similar manner, use the down arrow key to create text boxes in the Latent Variables
column and enter the names Visual and Verbal. Note that a label name, which may include
blanks, may not exceed 8 characters.

When done with the assignment of names, click the Next button to go to the Data dialog
box.
Select Correlations from the Statistics from: drop-down list box and also Correlations
from the Matrix to be analyzed drop-down list box. In the Number of Observations string

field type 145. For File type, select External ASCII Data and use the Browse button to
locate exs.cor in the splex folder. Click OK when done.
2.3 Drawing the Path Diagram
We now proceed with the actual drawing of the path diagram. Start by clicking on the VIS
PERC label under the Observed variables portion of the Labels window. Hold the mouse
button down and "drag" the label to the draw area indicated by the grid lines. A
rectangular-shaped object will appear on this part of the screen when the mouse button is
released as shown below. Note also that
 The observed variables are assumed to be X (or independent) variables unless

appropriate squares under the Y-column are clicked.
 The latent variables are assumed to be KSI (or independent) variables unless
appropriate squares under the Eta-column are clicked.
Repeat the same procedure for each of the remaining 5 variables by clicking on a label
(left mouse button) and dragging the object to the draw area on the screen. The result of
these operations should look similar to the image given below:

The rectangles representing the six observed variables can be properly aligned by going
through the following steps:
o Choose Select all from the Edit menu, or draw a rectangle with the mouse
pointer around all the objects to be included.
o Move the mouse pointer to the vicinity of the selected objects and click the
right mouse button. The menu shown below will appear on the draw part of
the screen.
o Select the Align>Left option. Once this is done, select the Even Space >
Vertically option. One can, alternatively, use the Align option on the Image
menu to achieve this.

The rectangular objects will now be aligned and one can proceed by dragging the latent
variables, Visual and Verbal to the path diagram window. Note that latent variables are
represented by elliptically shaped objects.
Finally, arrows can be drawn pointing from the latent variables to the observed variables.
To accomplish this, click on the single-headed arrow of the Draw toolbar (seen on the
right of the picture given below) and move the mouse pointer to within one of the
elliptically shaped objects. With the left mouse button held down, drag the arrow to
within a rectangular-shaped object. Release the mouse button when the colors of both
objects change. Note that the Draw menu on the main menu bar can also be used when
drawing path diagrams. Related draw tool options are contained in this menu.
Proceed in a similar fashion to graphically display the relationships between the observed
and latent variables as shown on the next image.

Select the Build SIMPLIS Syntax option from the Setup menu. SIMPLIS syntax (not shown
below) will be built from the path diagram and is stored in a system file named cfa6.spj.
Click the Run LISREL icon button to fit the CFA model to the data. Parameter estimates
are shown on the path diagram given below.

By modifying the path diagram, model specifications may be changed. The covariance
between Visual and Verbal may, for example, be fixed at a value of 0.5 by clicking (right
mouse button) on the two-headed arrow representing this covariance. This will enable the
pop-up menu shown below. Select the Fix option using the left mouse button. Once done,
select the Set Value… option.

Click the Run LISREL icon button to fit the revised CFA model to the data. The path
diagram,  2 -fit statistic and RMSEA are shown below.
3 Multi-Sample Analyses Using Path Diagrams
In the previous sections we showed how LISREL models may be specified and estimated
by drawing the appropriate path diagram from which the SIMPLIS or LISREL syntax is
created. The examples used were based on data from a single sample. The path diagram
can, however, also be used to create syntax for models based on data from several
samples simultaneously, according to a multiple-group LISREL model with some or all
parameters constrained to be equal over groups.
Consider a set of G mutually exclusive groups of individuals. It is assumed that a number

of variables have been measured on a number of individuals from each of these
populations. This approach is particularly useful in comparing a number of treatment and
control groups regardless of whether individuals have been assigned to the groups
randomly or not.
Any LISREL model may be specified and fitted for each group of data. However, LISREL
assumes by default that the models are identical over groups, i.e., all relationships and all
parameters are the same in each group. Thus, only differences between groups need to be
specified.
3.1 Testing Equality of Factor Structures
Table 1 below gives sample covariance matrices for two samples ( N1 = 865, N 2 = 900,
respectively) of candidates who took the Scholastic Aptitude Test in January 1971. The
four measures are, in order, VERBAL40, a 40-item verbal aptitude section, VERBAL50, a
separately timed 50-item verbal aptitude section, MATH35, a 35-item math aptitude
section, and MATH25, a separately timed 25-item math aptitude section.
Table 1: Covariance Matrices for SAT Verbal and Math Sections
Covariance Matrix for Group 1
Tests VERBAL40 VERBAL50 MATH35 MATH25

VERBAL40 63.382
VERBAL50 70.984 110.237
MATH35 41.710 52.747 60.584
MATH25 30.218 37.489 36.392 32.295
Covariance Matrix for Group 2
VERBAL40 VERBAL50 MATH35 MATH25

Tests
VERBAL40 67.898
VERBAL50 72.301 107.330
MATH35 40.549 55.347 63.203
MATH25 28.976 38.896 39.261 35.403
The data are used here to illustrate how one can test equality of factor loadings and factor
correlations in a confirmatory factor analysis model, while allowing the error variances to
be different.
We regard VERBAL40 and VERBAL50 as indicators of a latent variable Verbal and MATH35
and MATH25 as indicators of a latent variable Math. The model we consider is shown in
the figure below.

There are three sets of parameters in the model:
o the four factor loadings corresponding to the paths from Verbal and Math to the
observed variables,
o the correlation between Verbal and Math, and
o the four error variances of the observed variables.
We begin by constructing a data file pdex10.cov containing the two covariance matrices.
In free format, each covariance matrix can be written on one line. The file pdex10.cov
(not part of the LISREL installation) is given below and once created, is stored in the
tutorial folder.
We assume that all parameters are the same in both groups. The syntax file pdex10a.spl
is:
Group 1: Testing Equality Of Factor Structures Model A: Factor Loadings, Factor Correlation,
Error Variances Invariant
Observed Variables: VERBAL40 VERBAL50 MATH35 MATH25
Covariance Matrix from File PDEX10.COV
Sample Size = 865
Latent Variables: Verbal Math
Relationships: VERBAL40 VERBAL50 = Verbal
MATH35 MATH25 = Math
Group 2: Testing Equality Of Factor Structures
Covariance Matrix from File PDEX10.COV
Sample Size = 900
Set the error variances of VERBAL40-MATH25 free

End of Problem
In the syntax above, the data in the covariance matrices and the sample sizes are
different. The names of the variables, observed as well as latent, are the same, and the
model is the same.
Select the New option from the File menu and select Path Diagram in the New dialog box.
Click OK when done. Select the tutorial folder and enter the name pdex10.pth. When
done, click Save.
It is possible to control specific SIMPLIS default features such as Maximum Number of

Iterations or Method of Estimation. From the Output menu, select SIMPLIS Outputs.
The SIMPLIS Outputs dialog box will appear. Customize this dialog box according to
your preferences, for example, by changing the Number of Decimals in the Output option
to 4 and by making sure that the Invoke path diagram option is selected. Click OK when
done.

Before drawing the path diagram, select Toolbars from the View menu and ensure that the
items shown below are selected. These are: Toolbar, Status Bar, Typebar, Variables,
Drawing Bar. Optionally, check the Grid Lines option.
Select the Title and Comments option from the Setup menu to obtain the Titles and
Comments dialog box.
The first item on the Setup dialog box is the Title for the particular problem. Provision is
also made for any additional Comments that the user may wish to enter. After typing in
the title and (optionally) the comments, click Next to go to the Group Names dialog box.

Next we enter a description for each group. Group labels can be inserted by clicking on
the first string field, entering the label for the first group and then using the down arrow
on the computer keyboard to create the next group’s string field. When done, click Next
to go to the Labels dialog box.
The default number of variables shown on the Labels dialog box is 3, these being CONST,
VAR 1 and VAR 2. Move the mouse pointer to the Observed Variables box and click in the
string field of CONST. Rename this variable to VERBAL40. Press the "down arrow" on the
keyboard to move to the second observed variable string field and enter the label
VERBAL50. Proceed in a similar way to enter the labels MATH35, MATH 25 and the labels
Verbal and Math for the latent variables. Click Next when done to go to the Data dialog
box.

From the Data dialog box, select Covariances from the Statistics from: drop-down list box
and also Covariances from the Matrix to be analyzed drop-down list box. Enter 865 in the
Number of Observations string field. For File type select External ASCII Data.

Click the Browse button to select the path and name of the file that contains the values of
the covariance matrices of the 2 groups. For the present example, the folder is tutorial and
the file is pdex10.cov, which was created earlier in this section. Click Open when done.
From the Data dialog box, select the second group from the Groups: drop-down list box.
Proceed as outlined above to enter the number of observations (900), file type and matrix
to be analyzed.
3.2 Drawing the Path Diagram
We now proceed with the actual drawing of the path diagram. Start by clicking on the
VERBAL40 label under the Observed variables portion of the Variables window. Hold the
mouse button down and drag the label to the draw area. A rectangular-shaped object will
appear on this part of the screen when the mouse button is released as shown below.
Repeat this procedure until all observed variables are dragged to the drawing area as
shown in the next image.
Note also that:
 The observed variables are assumed to be X (or independent) variables unless

appropriate check boxes under the Y-column are checked.
 The latent variables are assumed to be Ksi (or independent) variables unless
appropriate check boxes under the Eta-column are checked.
Proceed by dragging the latent variables, Verbal and Math, to the draw area of the path
window. Note that the latent variables are represented by elliptically shaped objects.
Finally, arrows can be drawn pointing from the latent variables to the observed variables.
To accomplish this, click on the single-headed arrow of the Draw toolbar (seen on the
right of the picture given below). As an alternative, use the Draw menu on the main menu
bar. The related drawing tool options may be selected from this menu.
Once the single-headed arrow (or One-way Path option) has been selected, move the
mouse pointer to within one of the elliptically shaped objects. With the left mouse button
held down, "drag" the arrow to within a rectangular-shaped object. When the colors of
both objects change (see diagram below) release the mouse button.
Proceed in a similar way to graphically display the relationships between the observed
and latent variables:
Select Build SIMPLIS Syntax from the Setup menu. SIMPLIS syntax will be built from the
path diagram and is stored in a system file named pdex10.spj shown below.

Click the Run LISREL icon button to fit the multi-sample model to the data. The
parameter estimates and  2 goodness-of-fit index are shown on the path diagram
displayed below:

The values shown on the path diagram are the parameter estimates. This is the default
selection. It is, however, also possible to view the t-values, modification indices or
expected changes by selecting the View menu on the main menu bar. Alternatively, use
the drop-down list by clicking in the Estimates: pane as shown below. To view t-values,
select the T-values option. The figure below shows how this is done and also indicates the
t-values on the path diagram.
To view the corresponding t-values for the second group, select group 2 on the menu bar.
The resultant changes in the path diagram are shown in the next image.
In the example above, the error variances, factor loadings and factor covariance are
constrained to be equal (invariant) across the two groups. To test for invariance of factor
loadings and covariance only, the error variances can be freed as follows. Make sure that
the path diagram is displayed, and then select Group 2 from the Groups: text box. Right-
click in the error variance of VERBAL40 and click the Free option. Repeat this procedure
for the remaining three error variances, and then select the Setup, Build SIMPLIS Syntax
option.

The actions described above result in a new SIMPLIS project file that contains the syntax
required to free the error variances across groups as shown below.
When done, click the Run LISREL icon. The output file reveals two solutions, one for
each group. The value of  2 is reported only after the second group.

In a multi-sample analysis, the  2 is a measure of fit of all models in all groups, and, in
general, this  2 cannot be decomposed into a  2 for each group separately. For our

example,  2 is 10.87 with 7 degrees of freedom (p = 0.14), so that the model does fit the
data.
4 Analysis of Ordinal Data using an asymptotic

covariance matrix.
Aish and Jöreskog (1990) analyzed data on political attitudes. Their data consist of six
ordinal variables measured on the same people at two occasions. Labels assigned to the
12 variables are as follows: NOSAY1, VOTING1, COMPLEX1, NOCARE1, TOUCH1,
INTERES1 (first occasion) and NOSAY2, VOTING2, COMPLEX2, NOCARE2, TOUCH2 and
INTERES2 (second occasion). These data are saved in the LISREL system data file
panelusa.lsf in the tutorial folder.
In this example, the following concepts are illustrated:
o calculation of polychoric correlations and asymptotic covariances

o drawing a path diagram for a given model
o creation of variable names for the path diagram
o creating LISREL and or SIMPLIS syntax from the path diagram
o the basic model and its components
o estimation types and the conceptual path diagram
o saving the path diagram in graphics format.
4.1 Calculation of Correlations and Asymptotic Covariance

Matrix
Select the Open option from the File menu to obtain the Open dialog box. Select LISREL
Data (*.lsf) from the Files of type: drop-down list box. Select panelusa.lsf from the
tutorial folder and click Open when done. The PRELIS system file (*.lsf) is displayed in
the form of a spreadsheet as shown below:
Select Output Options from the Statistics menu to obtain the Output dialog box. The
Output Options dialog box allows the user to save various matrices as well as the raw
data as files. Select Correlations (to be saved as a LISREL System File (also referred to as
a LISREL summary data file) with extension *.dsf) from the Moment Matrix drop-down list
box. Check the LISREL System Data check box. Also check the Save to File check box
below Asymptotic Covariance Matrix and enter the name panelusa.acm in the string field.
With the required selections made, click OK to run PRELIS. Note that PRELIS computes
polychoric correlations for all variables that are defined as ordinal, provided that the
number of distinct values of the variables is less than 15.
A portion of the PRELIS output is given below:

4.2 Drawing a Path Diagram
In this example we use PRELIS to compute the polychoric correlation matrix and
corresponding asymptotic covariance matrix of these correlations. The next step is to
create a path diagram from which the LISREL or SIMPLIS syntax may be built.
Select the New option from the File menu to obtain the New dialog box. From the New
dialog box, select Path Diagram and click OK when done.
The Save As dialog box will appear. Save the new path diagram as panelusa.pth in the
tutorial folder.
Before proceeding to the next step, ensure that the Toolbar, Status Bar, Type Bar,
Variables and Drawing Bar options are selected, see below.
Select the Title and Comments option from the Setup menu to obtain the Title and
Comments dialog box. Use the Title and Comments dialog box to enter a title and to add
additional comments as shown in the example below:

Click Next to go to the Group Names dialog box. Since the present example is based on
the analysis of a single group, we can leave this dialog box blank and go to the variable
names (Labels) dialog box by clicking Next.
4.3 Reading Variable Names from the DSF File

The data system file panelusa.dsf, which was created during the PRELIS analysis,
contains the variable names, the location of the asymptotic covariance matrix, and the
matrix of polychoric correlation coefficients. The variable names can be read in by
clicking Add/Read Variables on the Labels dialog box.
In doing so, the Add/Read Variables dialog box is activated and on this dialog box click
the Read from File: LISREL Summary File radio button.

Click Browse to select the file panelusa.dsf from the tutorial folder. Click Open to return
to the Add/Read Variables dialog box.
Click OK to obtain the list of twelve names shown in the Labels dialog box below. To
complete the information in this dialog box, add the names of the two latent variables
Efficac1 and Efficac2 in the Latent Variables window. See the notes at the bottom of the
Labels window. To start typing a name, click anywhere within the Latent Variables
window and type in the name. Alternatively, use the Add Latent Variables button.
Once names for the observed and latent variables are entered, click Next to go to the Data
dialog box. Select Correlations from the Statistics from drop-down list box.
Note:
Since the location of the asymptotic covariance matrix is stored in the *.dsf file, it is not
necessary to enter the weight matrix details on the Data dialog box.

At this point, one may view the contents of the panelusa.dsf file by clicking the Edit
button. A portion of the correlation matrix is shown below. Note that only the lower
triangular halve contains correlation values.
Close the spreadsheet to return to the path diagram environment. The list of variable
names will appear on the left side on the computer screen. Define the variables NOSAY1,
VOTING1, COMPLEX1, NOSAY2, VOTING2 and COMPLEX2 as Y-variables by checking the
appropriate boxes under the heading Y. Also check the Eta-variable boxes to define
Efficac1 and Efficac2 as Eta latent variables.
The observed values to be used in the drawing of the path diagram can be dragged to the
path diagram area by clicking on a variable name (mouse left button). By holding this
button down, a variable can be dragged to a desired position.

Drag the two latent variables to the path diagram window. To create paths between
variables we proceed as follows: Click on the single headed arrow icon button on the
Draw toolbar as shown. Move the mouse cursor to the middle of the Efficac1 ellipse, and
then click the left mouse button. Hold the button down while dragging the arrow to
within the NOSAY1 rectangle. The colors of both objects should change. Repeat the
procedure to add the remaining paths as illustrated:

4.4 Creating LISREL and/or SIMPLIS Syntax from the Path
Diagram
Prior to building the SIMPLIS or LISREL syntax, we adjust the default number of decimals
from 2 to 4. This is accomplished by selecting the Output, LISREL outputs, Estimations…
option.
The method of estimation is also changed from Maximum Likelihood to Weighted Least
Squares by selecting the Output, LISREL output, Estimations option.

Select Build LISREL Syntax from the Setup menu to obtain the following LISREL project
window that displays the contents of the LISREL syntax file panelusa.lpj.
Click the Run LISREL icon button to start the analysis. The path diagram shows the
estimated parameter values and  2 -value.

Often researchers prefer to set one factor loading each set that are indicators of a specific
latent variable (factor) equal to one. This can be accomplished by editing the contents of
the .lpj file or by right clicking on a factor loading, and then fixing its value to one. The
resultant .lpj file is shown below.
The corresponding SIMPLIS input syntax can be built from the path diagram obtained by
running LISREL. This is accomplished by selecting the Build SIMPLIS Syntax option from
the Setup menu.

A selection of the SIMPLIS syntax (panelusa.spj) is displayed in the SIMPLIS project
window as shown in the next image.
This syntax produces a model that consists of X-variables and Y-variables, a KSI latent
variable and an Eta latent variable. Click the Run LISREL icon button to run the analysis.
It is interesting to note that the Y-only model and the model shown below are equivalent.

4.5 The Basic Model and its Components
The model shown above is called the Basic Model, which can be divided into the X-Model,
the Y-Model and the Structural Model. Each of these models may be viewed separately by
clicking on the Basic Model drop-down list box to obtain the list of models.
By selecting the X-Model model type, this portion of the path diagram is displayed. The
X-Model selection and Y-Model selection are shown below.

The selection of the Structural Model option produces that part of the path diagram:
Finally, the selection of the correlated errors option produces the following diagram.
Note that use was made of the two-headed arrow to add the error covariance paths to the
diagram.

When the Run LISREL icon button is clicked, the following path diagram is obtained.
From this display, it is seen that the  2 -value has significantly decreased as a result of
the addition of the error covariances.

4.6 Estimation Types and the Conceptual Path Diagram
LISREL also allows the user to view the parameter Estimates, Completely Standardized
Solution, Conceptual Diagram, T-values, etc. by clicking on View (main menu bar) or by
clicking on the Estimates drop-down list box as shown below:
The selection of the Conceptual Diagram option produces a path diagram that contains
only variable names and paths between variables. A black and white display of the path
diagram can be obtained by selecting the Black and White option from the Image menu.
4.7 Saving the Path Diagram in Graphics Format
The conceptual path diagram, or one with an alternative selection from the Estimations
bar, can be saved as a Windows Metafile (*.wmf) or a Graphic Interchange Format (*.gif)
file, which may be included as a graphic in, for example, Microsoft Word or Powerpoint.
To save as a *.wmf file, select the Export as Metafile option from the File menu.
5 Empty Path Diagrams

5.1 Motivation
Empty path diagrams can be used in teaching to discuss various types of models. The
commands for creating empty path diagram are explained here. After the empty path
diagram has been produced, user can start drawing the paths (arrows).
5.2 Examples
Example 1
If no relationships are specified, LISREL treats all observed variables as x-variables and all
latent variables as Ksi-variables.
Raw Data from File NPV.LSF

Latent Variables: Visual Verbal Speed
Path Diagram
End of Problem
The LISREL model is (omitting intercepts)
x  Λ xξ  δ
Example 2
This example illustrates a model with only x- and y-variables.
Observed Variables: Y1 - Y3 X1 X2
Covariance matrix:
14.610
-5.250 11.017
-8.057 11.087 31.971
-0.482 0.677 1.559 1.021
-18.857 17.861 28.250 7.139 215.662
Sample Size 173
Y-Variables Y1-Y3
Path Diagram
End of problem
The LISREL model is (omitting intercepts)
y  By  Γx  ζ

Example 3
The next example illustrates a model with y-, Eta- and Ksi-variables (but no x-variables
and omitting intercepts)
Raw Data from File NPV.LSF

Latent Variables: Visual Verbal Speed Ksi
Y-Variables ’VIS PERC’ - SCCAPS
Eta-Variables Visual - Speed
Path Diagram
End of Problem
The LISREL model is
y  Λyη  ε
η  Γξ  ζ
Example 4
The last example illustrates a model with only y- and Eta-variables (omitting intercepts)
Observed Variables: Y1-Y7

Correlation Matrix
(20F4.3) Symmetric
1000
5561000
456 4901000
439 445 5621000
415 418 496 5121000
399 383 456 469 5511000
387 364 445 442 500 5441000
Sample Size 1600
Latent Variables Eta1-Eta7
Y-Variables Y1-Y7
Eta-Variables Eta1-Eta7
Path Diagram
End of Problem
The LISREL model is
y  Λyη  ε
η  Bη  ζ

6 Using SIMPLIS Project Files
LISREL allows the user to create SIMPLIS syntax by making use of a dialog box keypad.
This keypad, see illustration below, is operated by clicking (left mouse button) on the
appropriate symbol. Note that the <== symbol represents the backspace key while the
symbol <--| denotes the enter key. The path diagram produced by running the program
can subsequently be modified to change the model specifications.
As an illustration, we consider path analysis for latent variables, which, in its most
general form, is a structural equation system for a set of latent variables classified as
dependent or independent. In the application described below, the system is recursive
(see the LISREL 8: SIMPLIS command language guide.)
Recursive models are particularly useful for analyzing data from longitudinal studies in
psychology, education and sociology. The characteristic feature of a longitudinal research
design is that the same measurements are used on the same people at two or more
occasions. Wheaton, et al. (1977) reported on a study concerned with the stability over
time of attitudes such as alienation, and the relation to background variables such as
education and occupation. Data on attitude scales were collected from 932 persons in two
rural regions in Illinois at three points in time: 1966, 1967, and 1971. The variables used
for the present example are the Anomia subscale and the Powerlessness subscale, taken to
be indicators of Alienation. This example uses data from 1967 and 1971 only.
The background variables are the respondent’s education (years of schooling completed)
and Duncan’s Socio-economic Index (SEI). These are taken to be indicators of the
respondent’s socio-economic status (Ses). The sample covariance matrix of the six
observed variables is given below:
y1 y2 y3 y4 x1 x2
ANOMIA67 11.834
POWERL67 6.947 9.364
ANOMIA71 6.819 5.091 12.532
POWERL71 4.783 5.028 7.495 9.986
EDUC -3.839 -3.889 -3.841 -3.625 9.610
SEI* -2.190 -1.883 -2.175 -1.878 3.552 4.503
*The variable SEI has been scaled down by a factor of 10.

The path diagram for this model is shown below:
Notes:
o The error terms of ANOMIA and POWERL are specified to be correlated over
time.
o The four one-way arrows to the right of the diagram represent the
measurement errors in ANOMIA67, POWERL67, ANOMIA71 and POWERL71
respectively.
o The two-way arrows on the right indicate that some of the measurement errors
are correlated. The covariance between the two error terms for each variable
can be interpreted as a specific error variance. For other models for the same
data, see Jöreskog & Sörbom (1989, pp. 213-223).
To set up this model for SIMPLIS is straightforward as shown in the following syntax file:
Stability of Alienation
Observed Variables
ANOMIA67 POWERL67 ANOMIA71 POWERL71 EDUC SEI
Covariance matrix
11.834
6.947 9.364
6.819 5.091 12.532
4.783 5.028 7.495 9.986
-3.839 -3.889 -3.841 -3.625 9.610
-2.190 -1.883 -2.175 -1.878 3.552 4.503
Sample Size 932
Latent Variables Alien67 Alien71 Ses
Relationships
ANOMIA67 POWERL67 = Alien67
EDUC SEI = Ses
Alien67 = Ses
Alien71 = Alien67 Ses
Set Error Covariance of ANOMIA67 and ANOMIA71 Free
Set Error Covariance of POWERL67 and POWERL71 Free
Path Diagram
End of Problem
Notes:
 The model is specified in terms of relationships. The first three lines under
Relationships specify the relationships between the observed and the latent
variables. For example, ANOMIA71 POWERL71 = Alien71 means that the observed
variables ANOMIA71 and POWERL71 depend on the latent variable Alien71.
 The Line Alien71 = Alien67 Ses means that the latent variable Alien71 depends on
the two latent variables Alien67 and Ses. This is one of the structural relationships.
6.2 Building the SIMPLIS Syntax
The files ex6a.spl and ex6b.spl in the splex folder both contain the numeric values of the
covariance matrix of the observed variables.
From the File menu, select Open and locate ex6a.spl in the splex folder. Use the mouse
cursor (left button down) to mark the covariances as shown below. Use the Edit menu and
click Copy or, alternatively, press the Ctrl - C keys on the keyboard.
From File, New select Syntax Only and click OK. Use the Edit, Paste function to copy the
contents of the clipboard to the syntax window or, alternatively, use the Ctrl - V keys on
the keyboard. Use the File, Save As option to save the covariances to the file
wheaton.cov.

Select the New option from the File menu, and select SIMPLIS Project in the New dialog
box. Click OK to open the Save As dialog box.
Select a filename and a folder in which the SIMPLIS project should be saved. For the
present example the folder chosen is splex and the filename selected is SIMPLIS6.spj.
When done, click Save to go to the SIMPLIS project (SPJ) window.
Before generating syntax, you may optionally select SIMPLIS Outputs from the Output
menu. Customize this dialog box according to your preferences; for example, change the
Number of Decimals in the Output option to 4. When done, click OK.
The following step is to provide information regarding Title and Comments, Groups,
Variables, and Data. Select the Titles and Comments option from the Setup menu. The
first item on the Title and Comments dialog box is the title for the particular problem.
After typing in the title and (optionally) the comments, click on Next to go on with the
setup.

Since the present example is based o n the one group only, nothing is entered in the space
allowed for Group Labels, and Next is clicked to go to the Labels dialog box.
The default number of variables shown on the Labels dialog box is 3, these being CONST,
VAR 1 and VAR 2. Click in the CONST textbox and change the name to ANOMIA67. Enter
the label POWERL67 in the VAR 1 textbox and ANOMIA71 in the VAR 2 text box. Move the
mouse cursor into the ANOMIA71 textbox and use the down arrow to create a fourth text
box. Type in POWERL71 and repeat the procedure for EDUC and SEI.
In a similar manner, use the down arrow key to create text boxes in the Latent Variables
column and enter the names Alien67, Alien71 and Ses. Note that a label name, which may
include blanks, may not exceed 8 characters.

On the Data dialog box select Covariances from the Statistics from: drop-down list box.
Type 932 in the Number of Observations string field. For File type, select External ASCII
data and use the Browse button to select wheaton.cov.
6.3 Building and Completing the SIMPLIS Syntax
Select Build SIMPLIS Syntax from the Setup menu to create the basic SIMPLIS syntax. The
Build SIMPLIS Syntax option on the Setup menu provides the user with a skeleton of
SIMPLIS commands. There are a few commands that must be included to complete the
SIMPLIS syntax. The first of these commands is:
ANOMIA67 POWERL67=Alien67
To add a line in the syntax window, move the mouse cursor to the end of the keyword
Relationships and click once. Now move the cursor to the Enter key ( <--| ) on the keypad
and click to insert a blank line below the Relationships line.
To enter the line
proceed as follows:
 Click on the observed variable ANOMIA67 (First variable under the heading
Observed) and with the mouse button held down drag the variable name to the
syntax window.
 Repeat by dragging the variable name POWERL67 next to ANOMIA67. To continue,
click on the "=" key and then drag the latent variable name Alien67 (first variable
under the heading Latent).
To complete, move the mouse cursor to the end of the line
and click. Then click on the enter key and repeat the steps outlined above to insert the
lines:

EDUC SEI = Ses
Alien67 = Ses
Alien71 = Alien67 Ses
Finally, enter the commands:
Set Error Covariance of ANOMIA67 ANOMIA71 free

and
Set Error Covariance of POWERL67 POWERL71 free
by using the keypad and by dragging the variable names to the appropriate positions. The
completed syntax should correspond to the lines shown in the SPJ window below:
To run SIMPLIS, click the Run LISREL icon button on the main menu bar.
A path diagram as shown below is produced, as well as the usual output file.

7 Using LISREL Project Files
LISREL allows the user to create LISREL syntax interactively by making use of a set of
menus. As an illustration, we consider a second-order factor analysis model.
Second-order factor analysis
The following equation
y  Λyη  ε
is in the form of a factor analysis model for y with first order factors η and
measurement errors ε . Now suppose that the variables η in turn can be accounted for by
a set of so called second-order factors ξ , so that
η  Γξ  ς
where Γ is a matrix of second-order factor loadings and ς is a vector of unique

components for η .
To illustrate the model, we use data on some cognitive ability tests. The standard
deviations and correlations of two forms of each of five tests are given in the table below.
The sample size (N) is 267.
Table 2: Correlations and Standard Deviations for Some Cognitive Tests
GESCOM-A 2.42 1
GESCOM-B 2.80 0.74 1
CONWORD-A 3.40 0.33 0.42 1
CONWORD-B 3.19 0.34 0.39 0.65 1
HIDPAT-A 1.94 0.26 0.21 0.15 0.18 1
HIDPAT-B 1.79 0.23 0.24 0.22 0.21 0.77 1
THIROUND 5.63 0.15 0.12 0.14 0.11 0.17 0.20 1
THIBLUE 3.10 0.14 0.14 0.14 0.15 0.06 0.09 0.42 1
VOCABU_A 3.05 -.04 0.03 0.09 0.16 0.06 0.09 0.19 0.21 1
VOCABU_B 2.25 0.02 0.02 0.10 0.23 0.024 0.07 0.09 0.21 0.72 1
We shall examine the hypothesis that the two forms of each test are tau-equivalent,
except for the two-word fluency tests Things Round and Things Blue that are only assumed
to be congeneric. The five true scores are postulated to depend on two factors, the first,
Speed Closure, being measured by the first three tests and the second, Vocabulary, being
measured by the last two tests. The model specification is:
Here * denotes parameters to be estimated and "0" and "1" are fixed parameters. The
LISREL syntax file for this model is as follows:
Second order factor analysis

DA NI=10 NO=267
LA FI=EX62.DAT
KM FI=EX62.DAT
SD FI=EX62.DAT
MO NY=10 NE=5 NK=2 GA=FI PH=ST PS=DI
LE
GESCOM CONWOR HIDPAT THINGS VOCABU
LK
SPEEDCLO VOCABUL
VA 1 LY 1 1 LY 2 1 LY 3 2 LY 4 2 LY 5 3 LY 6 3 LY 7 4 LY 9 5 LY 10 5
FR LY 8 4 GA 1 1 GA 2 1 GA 3 1 GA 4 2 GA 5 2
ST 1 ALL
OU SE TV SS NS
The goodness-of-fit statistic  2 is 53.06 with 33 degrees of freedom.
7.1 Building the LISREL Syntax
From the File menu, select New to obtain the New dialog box. Select LISREL Project and
click OK to obtain the Save As dialog box.
Select the tutorial folder and filename sofa.lpj. When done, click Save.
The following step is to provide information regarding the title, additional comments to
be included in the syntax file, the data file and the variables in it. This is accomplished by
selecting the Title and Comments option from the Setup menu. The Title and Comments
dialog box appears. Enter a title and (optional) comments in the Title string field and the
Comments text box.

Since the present example is based on the analysis of a single group, nothing is entered in
the space allowed for Group Labels, and Next is clicked to go to the Labels dialog box.
The default number of variables on the Labels dialog box is 2, these being VAR 1 and VAR
2. Since the present analysis calls for 10 variables, the mouse cursor is moved to the
Observed Variables box. By pressing the Insert key on the computer keyboard eight
times, provision is made for eight additional observed variables, and therefore the total
number becomes ten. Alternatively, the Add List option may be used and var3-var10
entered (or, for example, X3-X10). If the Add List option was used, click OK to return to
the Labels dialog box. Variable labels can alternatively be read in from an existing *.lsf
or *.dsf file. See the section on LISREL system files for more information on this option.
A label may be assigned to each of the 10 variables by clicking on the corresponding

button (numbered 1,2,3, … under the Observed Variables column). The background
color of the space allowed for entering the variable name will change and one can go
ahead and type the appropriate label. For the present example, the labels are GESCOM A,
GESCOM B, CONWOR A, CONWOR B, HIDPAT A, HIDPAT B, THIROUND, THIBLUE,
VOCABU A and VOCABU B respectively. Note that a label name, which may include
blanks, may not exceed 8 characters. Note also that the labels for the variables may
instead be entered by, for example, clicking on VAR 1. Type in GESCOM A and use the
down arrow on the keyboard to create the next space. Type GESCOM B and continue in a
similar fashion until all the labels are typed in.
Since we have 7 latent variables, Gescom, Conwor, Hidpat, Things, Vocabu, Speedclo and
Vocabu1, the procedure outlined above for the observed variables is repeated by selecting
Add List on the Latent Variables portion of the Labels dialog box. Add var1-var7 and click
on OK (alternatively the keyboard down arrow may be used). Assign labels to these latent
variables by clicking on the number of each variable. Finally, click Next to go to the Data
dialog box.

From the Data dialog box, select Covariances from the Statistics from: drop-down list
box and also Covariances from the Matrix to be analyzed drop-down list box. Type 267 in
the Number of Observations string field. For File type, select External ASCII Data and then
click Browse to go to the Browse dialog box.

Select the ASCII file sofa.cov from the tutorial folder. Click Open to return to the Data
dialog box, and then select Next to go to the Define Observed Variables dialog box.
The program will now produce the Define Observed Variables dialog box, on which the
user may select Y-variables and X-variables from the list of Variable Names. Since the
present application calls for Y-variables only, all of the variable names are selected as Y-
variables. This is achieved by sequentially clicking on each variable number box and then
on the Select as Y option. Alternatively, with the left mouse button down, drag from
variable number 1 to number 10, release the mouse button and click on the Select as Y
button.
Click Next to go to the Define Latent Variables dialog box.
The latent variables consist of 5 Eta-variables and 2 Ksi-variables. Click on the

appropriate square under Variable Names and then choose Select Eta for the Eta-variables
(the first 5) or Select Ksi for the Ksi-variables (the last 2 variables). Once this is done,
click Next to go to the Model Parameters dialog box.

The Model Parameters dialog box shown below allows the user to:
o specify the form and mode of the matrices listed,

o specify which elements are fixed and which are free, and
o assign values to the elements of each of the matrices.

Since some changes need to be made to the default specifications for the Lambda-Y
matrix, highlight the Lambda-Y Full Matrix, Fixed option by clicking on it. This action will
cause the words Full Matrix and Fixed to appear in the Matrix From and Matrix Mode fields
respectively. Proceed by clicking Specify to obtain the dialog shown below.
The initial dialog box will indicate that all the elements of Lambda-Y are fixed and equal
to zero. The steps required to obtain the pattern shown in the dialog box below
corresponding to the statements is given next.
VA 1.0 LY(1,1) LY(2,1) LY(3,2) LY(4,2) LY(5,2) LY(6,3) LY(7,4) LY(9,5) LY(10,5)
FR LY(8,4),
o Click on the first rectangle next to the variable description button (that is
under the heading Gescom) and use the computer keyboard to type in the
value of 1, then press the keyboard Enter button.
o Proceed in a similar fashion to enter 1 in the second row, and finally a 1 in the
rectangle corresponding to the row VOCABU B and column Vocabu.
o By clicking on the rectangle in the row THIBLUE and column Things and then
on the Free button, the color of the rectangle will change to indicate that
LY(8,4) is free.
Note that one may move the mouse pointer to a vertical divider line on the dialog box
below. Column widths can subsequently be adjusted by clicking the left button and
dragging the vertical line to the left or the right of its present position.

The input specifications require BETA to be a full matrix with all elements fixed at a value
of 0. Since this is the default, one can proceed to specify the pattern and values for the
GAMMA matrix.
Highlight the Gamma Full Matrix, Fixed option and click on Specify.

Use the dialog box shown above to specify the following statements
FR GA(1,1) GA(2,1) GA(3,1) GA(4,2) GA(5,2)

VA 1.0 GA(1,1) GA(2,1) GA(3,1) GA(4,2) GA(5,2)
This is accomplished by clicking on the rectangle that corresponds to the Eta-variable

Gescom and the Ksi-variable Speedclo. Use the computer keyboard to type in 1 and click
again to activate this rectangle. Then select the Free button to change the status of the
GA(1,1) parameter to free. Continue in a similar fashion to obtain the pattern shown
above.
Since the default settings for the parameter matrices Psi and Theta-Epsilon are correctly set
at the default values (see dialog box below), click on Next to go to the Constraints dialog
box.
In the present example, no constraints are imposed on the elements of the parameter
matrices and therefore the Next button is clicked to go to the Output Selections dialog
box.

The Output Selections dialog box enables the user to create an output file that contains
selected parts of the LISREL printout. Make the appropriate selections by clicking on the
boxes at the left of the different options. Once this is done, click Next to go to the final
dialog box, which is the Save Matrices dialog box.

Any of the matrices or statistics shown on the dialog box illustration below may be
selected by clicking on the boxes.
As an example, we selected Lambda-Y, Gamma and t-values. These files will have the
same filename as the LISREL project file but with different file extensions, these being
*.lys, *.gas, and *.tvs respectively.
Click OK to return to the LISREL project (LPJ) window. The LISREL syntax, saved in the
file sofa.lpj, is shown below:
Click the Run LISREL icon button on the main menu bar to run the problem.

The path diagram for this example is shown below and the output is saved to the file
sofa.out.
One may modify the path diagram by deleting, fixing or adding paths.
8 Analysis with Missing Data

8.1 Single-Group Analysis with Missing Data
Data on attitude scales were collected from 932 persons in two rural regions in Illinois at
three points in time: 1966, 1967 and 1971. The variables used for the present example are
the Anomia subscale and the Powerlessness subscale, taken to be indicators of
Alienation. This example uses data from 1967 and 1971 only.
The background variables are the respondent’s education (years of schooling completed)
and Duncan’s Socioeconomic Index (SEI). These are taken to be indicators of the
respondent’s socioeconomic status (SES).
Observed Variables: Anomia 67, Powerlessness 67, Anomia 71, Powerlessness 71,
Education, SEI
Latent Variables: SES, Alienation 67, Alienation 71
The model shown in the path diagram below was fitted by running ex64d.lis (in the lisex
folder).

By adding SI=filename to the OU line of the syntax file, the fitted covariance matrix may
be saved to an external file.
Subsequently, a data set of size 1500, with 15% of the values missing at random, was
simulated by regarding the fitted covariance matrix as the true population covariance
matrix. This data set is stored as the text file wmas.dat in the missingex folder. A
corresponding *.lsf file, containing the variable names, missing value code, number of
records and the data is stored as wmas.lsf in the tutorial folder. Select File, New, Path
Diagram and save the path diagram in the tutorial folder. In the dialog box below, we used
the name sgroup.pth.

Click Save when done and select the Title and Options option from the Setup menu. Enter
a title and optional comments. Click Next to proceed to the Group Names dialog box and
again click Next to proceed to the Labels dialog box.
Click the Add/Read Variables button and on the Add/Read Variables dialog box select
LISREL System data file from the Read from file drop-down list. Use the Browse button to
locate wmas.lsf in the tutorial folder.

Click OK when done and add the latent variable names as shown below. By default,
LISREL uses the file selected for reading in variable names as the data file. It is therefore
not necessary to proceed to the Data menu. Click OK to return to the path diagram
window.
Mark the first four variables as Y-variables and the first two latent variables as Eta-
variables. (Left) click on each variable name, and with the mouse button down, drag the
variable to the path diagram (PTH) window.

Start with the Y-variables first, then continue with the Eta-variables, then the Ksi-variables
and finally the X-variables. In other words, start with ANOMIA67 and end with SOCIOIND.
Click on the draw tool (single-headed arrow) and add all the paths as shown below. Start
by moving the mouse cursor to the center of ALIEN67, and, with the mouse button down,
drag the arrow to the center of ANOMIA67 and release the button.
Once all the single paths have been drawn, click on the double-headed arrow. Move the
cursor to the center of the error variance arrow of ANOMIA67, then, with the left mouse
button held down, drag to the center of the error variance arrow of ANOMIA71. Repeat this
procedure to allow for a covariance between POWER67 and POWER71.

Select Build LISREL syntax from the Setup menu to produce the sgroup.lpj file. This file
is displayed in the LISREL project (LPJ) window shown below.
The syntax files missex4a.lis and missex4b.spl in the missingex folder contain the
corresponding LISREL and SIMPLIS syntax for fitting the model. Missing values for this
analysis are indicated by a value of -9. A FIML analysis is carried out if the DA line in the
LISREL syntax contains the keyword MI=<value>. This is required if the data file is not a
LSF file:
DA NI=6 NO=1500 MI=-9

LA
ANOMIA67 POWER67 ANOMIA71 POWER71 EDUCATIN SOCIOIND
RA FI=WMAS.DAT
MO NY=4 NX=2 NE=2 NK=1 BE=SD PS=DI TE=SY

LE
ALIEN67 ALIEN71
LK
SES
FR LY(2,1) LY(4,2) LX(2,1) TE(3,1) TE(4,2)
VA 1 LY(1,1) LY(3,2) LX(1,1)
PD
OU SE TV MI ND=3 PT
Click the Run LISREL icon button to obtain a path diagram and output file. Once the data
are read, estimates are obtained of the means and covariances and a 2ln L value is
printed. This value is the 2ln L value for the unrestricted model.
Portions of the output file are shown below:
(i) –2*Log(likelihood) and percentage cases missing
The EM-algorithm is used to obtain estimates of the population means and covariances.
LISREL uses these values to obtain starting values for the maximum likelihood procedure.
Convergence is attained in 8 iterations, and at convergence 2ln L equals 35822.65. This
value is also referred to as the 2ln L value for the saturated model.

(iv) Global Goodness of Fit Statistics, Missing Data Case
-2ln(L) for the saturated model = 35822.647

-2ln(L) for the fitted model = 35828.671
Degrees of Freedom = 4
Full Information ML Chi-Square = 6.02 (P = 0.20)
Root Mean Square Error of Approximation (RMSEA) = 0.018
8.2 Multiple-groups with Missing Data
Sörbom (1981) reanalyzed some data from the Head Start summer program previously
analyzed by Magidson (1977). Sörbom used data on 303 white children consisting of a
Head Start sample (N = 148) and a matched Control sample (N = 155). The children were
matched on gender and kindergarten attendance, but no attempt had been made to match
on social status variables.
The variables used in Sörbom’s re-analysis were:
X 1 = Mother’s education (MOTHEDUC)

X 2 = Father’s education (FATHEDUC)
X 3 = Father’s occupation (FATHOCCU)
X 4 = Family income (FAMILINC)
Y1 = Score on the Metropolitan Readiness Test (MRT)
Y2 = Score on the Illinois Test of Psycholinguistic Abilities (ITPA)
The following issues were examined:
 Test whether X 1 , X 2 , X 3 and X 4 can be regarded as indicators of a single

construct Ses (socioeconomic status) for both groups. Is the measurement model
the same for both groups? Is there a difference in the mean of Ses between
groups?
 Assuming that Y1 and Y2 can be used as indicators of another construct Ability
(cognitive ability), test whether the same measurement model applies to both
groups. Test the hypothesis of no difference in the mean of Ability between groups.
 Estimate the structural equation:
Ability     Ses  z
 Is  the same for the two groups? Test the hypothesis   0 and interpret the
results.
A conceptual path diagram for the model fitted to the data is shown below.
The fitted covariance matrices obtained from ex16d.spl (in the splex folder) were used to
simulate a control group data set (sample size 550 and percentage missing 15%) and to
simulate a Head Start data set (sample size of 600 and percentage missing 10%).

Note:
To invoke the FIML procedure for the analysis of missing data, the following three
statements must be given in the SIMPLIS syntax in the order shown below.
Missing Value Code <value>

Sample Size: <nsize>
Raw data from file <filename>
The complete SIMPLIS syntax file is shown below.
FIML: Example 1: SIMPLIS syntax

Group = Control
Observed Variables: MOTHEDUC FATHEDUC FATHOCCU FAMILINC MRT ITPA
Missing Value Code –9
Sample Size: 550
Raw Data from File CONTROL.DAT
Latent Variables: Ses Ability
Relationships:
MOTHEDUC = CONST + 1*Ses
FATHEDUC - FAMILINC = CONST + Ses
MRT = CONST + 1*Ability
ITPA = CONST + Ability
Ability = Ses
Let the Errors of MOTHEDUC and FATHEDUC correlate
Group = Head Start
Missing Value Code –9
Sample Size: 600
Raw Data from File EXPERIM.DAT
Relationships:
Ses = CONST
Ability = CONST + Ses
Set the Error Variances of MOTHEDUC - ITPA free
Set the Variance of Ses free
Set the Error Variance of Ability free
Let the Errors of MOTHEDUC and FATHEDUC correlate
LISREL Output: ND=3
Path Diagram
End of Problem
If the data are stored in a LISREL system file (*.lsf), the Observed Variables, Missing Value
Code, Sample Size and Raw data from file control.dat can be replaced with the statement
Raw data from file control.lsf
If a *.lsf file contains missing data, the user should ensure that a global missing value
code is assigned. This can be done by using the Data, Define Variables option.
Subsequently, we illustrate how to build the SIMPLIS (or LISREL) syntax by drawing a
path diagram.

Select the New option from the File menu. On the New dialog box, select Path Diagram
and save the path diagram in the tutorial folder as mgroup.pth. These steps are described
in detail in earlier sections of this guide. From the Setup menu, select Title and Comments
and provide a title and any optional comments. When done, click Next to go to the Group
Names dialog box.
Use the instructions on the bottom of the Group Names dialog box to enter group names,
then click Next to go to the Labels dialog box.
The files control.lsf and experim.lsf containing the Group1 and Group2 data respectively,
are stored in the tutorial folder. Click the Add/Read Variables button, select LISREL
System File from the Add/Read Variables dialog box and use the Browse button to locate
the file control.lsf in the tutorial folder. Click OK when done to return to the Labels dialog
box. Use the Add Latent Variables button to insert the names Ses and Ability.
Click the Next button to go to the Data dialog box. Since a *.lsf file was selected in the
previous step, the Data dialog box shows the File type: as LISREL System Data File and the
filename as control.lsf. Use the Groups: drop-down list box to select Group2=Head Start.
Once this is done, use the Browse button to locate the file experim.lsf. Click Open to
return to the Data dialog box.

Click OK when done. Select the variables MRT and ITPA as Y-variables and Ability as an
Eta-variable. Drag variable names to the path diagram screen in the order MRT, ITPA,
Ability, Ses, MOTHEDUC, FATHEDUC, FATHOCC, and FAMILINC. Use the draw toolbar to
draw the paths as shown below.
A more detailed description of this process is given in preceding sections of this guide.

When the path diagram is completed, use the Build SIMPLIS syntax option from the Setup
menu to create the SIMPLIS syntax file which is displayed in the SIMPLIS project window,
as shown below.
Note that no relationships are given for Group2 under the Relationships keyword. This
implies that all parameters are constrained to be equal across groups.
Click the Run LISREL icon to create an output file and the path diagram shown below.
Since it is not realistic to assume equal error variances across groups, we can set each
error variance free by moving the mouse to the appropriate arrow (e.g. MRT  0.96).
Right click to obtain the menu below and select the Free option. Proceed in a similar way
to free the error variances of ITPA, MOTHEDUC, FATHEDUC, FATHOCC and FAMILINC as
well as the covariance (5.45) between MOTHEDUC and FATHEDUC.

Also right click on the arrow between Ses and Ability to free that path. When this is done,
rebuild the SIMPLIS syntax using the Build SIMPLIS syntax option on the File menu.
A portion of the SIMPLIS project file showing the resultant relationships for group 2 is
shown below.
Click the Run LISREL icon button to produce the path diagram shown below.
When the missing value code, sample size and raw data file information are read in, the
EM algorithm for estimating the means and covariances under the unrestricted model is
started. From this portion of the output a percentage of 14.61% missing cases are reported
for group 1 and 9.92% for group 2.
Note that the estimated means and covariances are used to obtain starting values for the
FIML procedure. In addition, a 2ln L value is reported for each group. This value is
minus two times the log likelihood value obtained when no restrictions are imposed on
means and covariance matrices. From the output it follows that the sum of the 2ln L
values for the groups equals 11358.595 + 12716.065 = 24074.660.
The FIML procedure converged in 6 iterations. Portions of the output file are given below.
(i) LISREL Estimates (Maximum Likelihood) for the control group
Number of Iterations = 6
Measurement Equations
MRT = 0.86*Ability, Errorvar.= 1.06 , R**2 = 0.20

(0.11) (0.084)
7.78 12.58
ITPA = 2.74*Ability, Errorvar.= 7.89 , R**2 = 0.26

(0.69)
11.41
MOTHEDUC = 1.14*Ses, Errorvar.= 12.70, R**2 = 0.093

(0.14) (0.85)
8.42 14.92
FATHEDUC = 1.07*Ses, Errorvar.= 6.15 , R**2 = 0.16

(0.098) (0.43)
10.90 14.26
FATHOCC = 1.01*Ses, Errorvar.= 0.96 , R**2 = 0.51

(0.055) (0.11)
18.35 8.39
FAMILINC = 0.78*Ses, Errorvar.= 0.74 , R**2 = 0.45

(0.047) (0.076)
16.65 9.81
(ii) Global Goodness of Fit Statistics, Missing Data Case
-2ln(L) for the saturated model = 24074.660

-2ln(L) for the fitted model = 24106.532
Degrees of Freedom = 20
Full Information ML Chi-Square = 31.872 (P = 0.0447)
Root Mean Square Error of Approximation (RMSEA) = 0.032
The FIML  2 is obtained as the difference between -2 ln L (24105.18) for the fitted model
and -2 ln L for the unrestricted model and equals 31.872.

9 Latent Variable Scores
9.1 Calculation of Latent Variable Scores
To obtain scores for the latent variables of a structural equation model, the following
three steps are required.
Step 1
If the data are not available as a LISREL system data file (*.lsf) yet, use the Import Data in
Free Format or Import External Data in Other Formats options from the File menu to create
a *.lsf file.
Step 2
Use this *.lsf file to compute the desired sample covariance or correlation matrix to
which the structural equation model should be fitted. This may be done by running the
appropriate PRELIS syntax file or by using the Output Options option from the Statistics
menu. Either action will result in the creation a LISREL data summary file (*.dsf). A
*.dsf file contains all the data information that LISREL requires to fit the structural
equation model to the data.
Step 3
In SIMPLIS syntax, use the line
System File from file <filename>.DSF
to specify the data to be analyzed and insert the command line
LSFfile <filename>.LSF
after the Relationships paragraph where <filename> denotes the path and name of the *.lsf
file. The System File command replaces the Observed Variables paragraph and the Sample
Size line.
In LISREL syntax, use the command line
SY = <filename>.DSF
to specify the *.dsf file and the
LS = <filename>.LSF
command line to specify the *.lsf file. The SY command replaces the LA and the DA
commands.

When done, run the syntax file by clicking the Run LISREL icon button. When the output
file is displayed, open the relevant *.lsf file using the Open option from the File menu.
The latent variable scores appear as the last set of columns of this file.
9.2 An Illustrative Example
Chow (2000) used five indicators of self esteem (Selfest) and four indicators of
depression (Depress) to formulate a measurement model for the latent variables Self
Esteem and Depression. A path diagram of this measurement model is shown below.
The nine indicators were observed for a random sample size of 230 college graduates.
The resulting raw data are listed in the PRELIS System File depression.lsf, located in the
tutorial folder. A portion of the spreadsheet is shown below.

The PRELIS syntax file depression.prl contains the required syntax to compute the sample
covariance and to create the desired LISREL Data Summary File depression.dsf. This file
is generated by selecting the Output Options option from the Statistics menu. The Output
dialog box is shown below. Select Covariances and check the LISREL system data box.
Click OK to run PRELIS.
The contents of depression.prl are listed below.
SY= Depress.lsf
OU MA=CM XM XB XT
The SIMPLIS syntax file depress.spl in the tutorial folder is used to compute the latent
variable scores for the latent variables Self Esteem and Depression. The contents of this
file are listed below.
If LISREL is executed, a LISREL system data file depressionnew.lsf is created with the
latent variable scores as shown below.
The LISREL syntax file corresponding to the SIMPLIS syntax file depress.spl is shown
below.

10 Multilevel Confirmatory Factor Analysis
The next example illustrates
o importing of external data such as SPSS *.sav files

o drawing a multiple-group path diagram
o building SIMPLIS/LISREL syntax from a path diagram
o fitting a multilevel structural equations model to the data
The data set used in this example forms part of the data library of the Multilevel Project
at the University of London, and comes from the Junior School Project (Mortimore et al,
1988). Mathematics and language tests were administered in three consecutive years to
more than 1000 students from 50 primary schools, which were randomly selected from
primary schools maintained by the Inner London Education Authority. The data set is
stored in the tutorial folder as an SPSS for Windows file named jsp2.sav.
It is convenient to work with LISREL system data files (*.lsf) since LISREL can read the
variable names, the number of variables, the number of records and other relevant
information directly from these files. It is therefore advantageous to convert files from
other software packages to the *.lsf format.
10.1 Import of External Data

Use the Import External Data in Other Formats option from the File menu to obtain the
Input Database dialog box. Select SPSS for Windows (*.sav) from the Files of type drop-
down menu list and from the tutorial folder select jsp2.sav. Click Open to obtain the Save
As dialog box.
Enter the name jsp2.lsf in the File name string field. By clicking the Save button a LISREL
system data file is created and displayed in the form of a data spreadsheet.
Although no missing value codes were assigned in jsp2.sav, data values of -9 indicate
missing values and can be defined as such by selecting the Define Variables option from
the Data menu.
Click on MATH1 (or any other variable) to select it. Click Missing Values to obtain the
Missing Values dialog box. Enter the value -9.0 in the Global Missing Value edit box and
change the method of deletion from listwise to pairwise. The pairwise option ensures that
recoding and computing of variables do not change the number of cases.

10.2 Draw the Path Diagram
Click the OK buttons of the Missing Values and Define Variables dialog boxes to return to
the main window. Use the Save option from the File menu to save the changes to jsp2.lsf.
Use the New option from the File menu to open the New dialog box and select the Path
Diagram option from the New dialog box and assign a file name, for example,
msemex.pth. By clicking the Save button, the path diagram window appears.
Select the Title and Comments option from the Setup menu to open the Title and
Comments dialog box. Enter the title Multilevel CFA model for Numeric and Verbal Ability in
the Title string field to produce the following Title and Comments dialog box.
Click Next to open the Groups dialog box. Enter the label Group1: Between Schools in the
first string field. Use the down arrow key to create a string field for the second group.
Enter the label Group2: Within Schools in the second string field to produce the following
Groups dialog box.

Click Next to proceed to the Labels dialog box. Since the observed variables are stored in
the file jsp2.lsf, it is convenient to read the variable names from this file. Click Add/Read
Variables to open the Add/Read Variables dialog box. Select the LISREL System File
option from the Read from file drop-down list box.
Use the Browse button to select the file jsp2.lsf in the tutorial folder to produce the
following Add/Read Variables dialog box. Click OK to return to the Labels dialog box.
Click Add Latent Variables to open the Latent Variables dialog box. Enter the label
NABILITY for Numeric Ability to produce the following dialog box.

Click OK. Repeat the process to enter the label VABILITY for Verbal Ability. Click OK to
produce the following Labels dialog box.
Click OK to return to the path diagram window. Click and drag the observed variables
MATH1, MATH2, MATH3, ENG1, ENG2 and ENG3 one by one into the path diagram
window. Also click and drag the latent variables NABILITY and VABILITY one by one into
the path diagram window to produce the following display.
Use the one-directional arrow on the Draw toolbar to insert paths from NABILITY to
MATH1, MATH2 and MATH3 and also to insert paths from VABILITY to ENG1, ENG2 and
ENG3 as shown below.
Note:
Once this arrow is selected, move the mouse to within an ellipse representing a latent
variable. With the left mouse button down, drag the arrow to an observed variable.
Release the mouse button when the arrow is in the rectangle representing the observed
variable.

10.3 Build SIMPLIS/LISREL Syntax
Select the Build SIMPLIS Syntax option from the Setup menu to produce the text editor
window shown overleaf. Insert the command line $CLUSTER SCHOOL after the Raw Data
from File command:
Change the command line MATH1=NABILITY to MATH1=1*NABILITY to set the scale of

NABILITY. Change the command line ENG1=VABILITY to ENG1=1*VABILITY to set the scale

of VABILITY. Copy and paste the relationships of the Between Schools group to the
Relationships paragraph of the Within Schools group.
In the Relationships paragraph of the Within Schools group, insert the following SIMPLIS
commands
Set Variance of NABILITY Free

Set Variance of VABILITY Free
Set Covariance NABILITY VABILITY Free
10.4 Fit the Model to the Data
Click the Run LISREL icon button on the main toolbar to produce the path diagram for the
between schools group shown below:

LISREL produces an output file, msemex.out which contains the results of the analysis.
See the Additional Topics Guide for more information on multilevel structural equation
models. Examples are contained in the \msemex folder.

LISREL Examples Guide

Uploaded by

Copyright:

Available Formats

LISREL Examples Guide

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LISREL Examples Guide

Uploaded by

Copyright:

Available Formats

Table of Contents

1 Using Existing SIMPLIS Syntax Files ............................................................................ 3

LISREL Examples Guide 2

LIST OF SIMPLIS EXAMPLES IN THE \SPLEX FOLDER

Example 1: Regression of GNP Example 2: Prediction of Grade Averages

LISREL Examples Guide 3

LISREL Examples Guide 4

LISREL Examples Guide 5

The SIMPLIS Outputs dialog box is shown below:

LISREL Examples Guide 6

A LISREL project file (ex10a.lpj) is produced, containing the corresponding LISREL

LISREL Examples Guide 7

2 Using Existing LISREL Syntax Files

LISREL Examples Guide 8

Example 1: Hypothetical model estimated Example 3.1: Analysis of reader reliability

LISREL Examples Guide 9

LISREL Examples Guide 10

LISREL Examples Guide 11

2.1 Creating Syntax by Drawing Path Diagrams

Holzinger & Swineford (1939) collected data on twenty-six psychological tests

The path diagram for this model is given below:

The corresponding SIMPLIS syntax file is:

Six Psychological Variables-A Confirmatory Factor Analysis

LISREL Examples Guide 13

2.2 Title, Labels and Data

LISREL Examples Guide 15

LISREL Examples Guide 16

LISREL Examples Guide 17

2.3 Drawing the Path Diagram

 The observed variables are assumed to be X (or independent) variables unless

LISREL Examples Guide 18

LISREL Examples Guide 19

LISREL Examples Guide 21

LISREL Examples Guide 22

LISREL Examples Guide 23

3 Multi-Sample Analyses Using Path Diagrams

Consider a set of G mutually exclusive groups of individuals. It is assumed that a number

3.1 Testing Equality of Factor Structures

Table 1: Covariance Matrices for SAT Verbal and Math Sections

Covariance Matrix for Group 1

Tests VERBAL40 VERBAL50 MATH35 MATH25

VERBAL40 VERBAL50 MATH35 MATH25

LISREL Examples Guide 25

LISREL Examples Guide 26

It is possible to control specific SIMPLIS default features such as Maximum Number of

LISREL Examples Guide 27

LISREL Examples Guide 28

LISREL Examples Guide 29

LISREL Examples Guide 30

3.2 Drawing the Path Diagram

Note also that:

 The observed variables are assumed to be X (or independent) variables unless

LISREL Examples Guide 32

LISREL Examples Guide 33

LISREL Examples Guide 34

LISREL Examples Guide 35

4 Analysis of Ordinal Data using an asymptotic

In this example, the following concepts are illustrated:

o calculation of polychoric correlations and asymptotic covariances

4.1 Calculation of Correlations and Asymptotic Covariance