ANALYSIS OF MOMENT STRUCTURE (AMOS)

Rules for working on AMOS:

Rule No.1: Variables are of two types: Observed and Latent
Observed Variables are those which are directly measured from respondents,
represented by in AMOS and Latent Variables are those which are inferred
indirectly and represented by
Different Names
Latent Variable / Factor / Construct
Observed Variables / Statement / Item
Rule No.2: Variables are of two types: Dependent Variable (DV) &
Independent Variable (IDV)
DV are also called as Endogenous Variables. Endo, Greek root, mean IN
Hence these variable will contain an arrow towards them
IDV are also called as Exogenous Variables. Ex, Greek root, mean OUT
Hence these variable will contain an arrow going out from them.
Rule No.3: Any of the or can be DV or IDV.

Sales Advt

Motivation Incentives

Performance Leadership

Motivation Leadership
Rule No.4: Every DV must contain an error term.

Error/
Residual

Sales Advt

Error/
Residual

Motivation Incentives

Error/
Residual

Performance Leadership

Error/
Residual

Motivation Leadership

Rule No.5: None of the can exist on canvas without a box. Every
requires either a (Called Zero order CFA) or another (Called Second
order)

Rule No.6: The error terms and at least one statement must have a Regression
Weight of 1 (To remove Un-identification Problem).
------------------------------------------------------------------------------------------------
Assumptions of AMOS:
1. Linear Relation only
2. Metric data only (Likert scale and Ratio measurement)- No categorical variables
3. Normality
4. Minimum number of statement in a construct
5. Sample Adequacy (Determined by KMO)

Basics of AMOS:
1. Works on SPSS datafile (.sav). First step is always attached a SPSS datafile in AMOS.
2. There should not be any missing values in the attached file. Handle them in SPSS, and
attach clean SPSS file in AMOS.
3. Graphical Representation + Statistical Calculations
 AMOS TOOL BAR Column A Column B Column C
A B C ROW Tool Meaning Tool Meaning Tool Meaning
Draw Draw Draw
1 Observed
Variable
unobserved
Variable
Latent
Variable

Add
Relation
2 (Regression)
Correlation Error term
in DV
List variable List variable
Figure
3 Caption
in
Model
in
attached Data file
Select Select
Deselect
4 One
Object
All
Objects
All

Make Move Delete

5 Clone Objects Objects
Change shape Rotate
Flip object to 90
6 of
Objects
indicators of
Latent var degree

Move Touch up
7 parameter
value
variable

Select
Analysis Calculate
8 SPSS
Data file Properties Estimates

Copy diagram
Go to
9 on Canvas
Output
Save

Object Drag properties Preserve

10 Properties from Object to
Object
Symmetries

Zoom an Area
11 Zoom in Zoom Out

Entire
12 diagram on
one screen
Best Fit Examine

Multi- group
13 Bayesian
Analysis
Print

Undo Redo Find

14

IMPORTANT PLUG-INS

Draw 2-sided arrows between selected constructs

Write the name in all circles i.e.

Unobserved variables
Output Mode

Input Mode

Working on Different
groups. Sort of Split file

Canvas

Choose which Regression

Output to see on Canvas

Brief Output of Chi sq test

Show all AMOS files of the folder

(where the attached file is saved)
Colors in AMOS:
1. Black Default color
2. Red Action will be actuated here
3. Blue Selected items
4. Green Destination
Errors in AMOS:
A) Pop-up Errors

1. There is more than One

Variable with same name.

2. Variable (which is present in

SPSS) datafile) is represented by
Ellipse.

3. Variable represented by
rectangle in diagram is not an
observed variable (i.e. not
present in SPSS Data file)
 B) Non Pop-up Errors

1. Un-identification
Error
After clicking on Calculate Estimates,
Calculations are done but…..

But, the Red icon is not Active,

indicating some error.

To see error, click on B9 icon i.e. View

Text (Go to Output)
 Must put Regression
Weight of 1 on any of
the Arrow.

2. Not positive Remove such statements as indicated by error terms in Output.

definite Sample
Covariance matrices
 Case-1 Case-2 Case-3
Dep Sales Ability Confidence
Indep Ability Confidence Hours
Mean dep 5766.46 4.06 3.92
Mean indep 4.06 3.92 11.81
R 0.767 0.544 0.826
R2 0.588 0.296 0.682
Intercept 1475.385 1.622 0.118
Unstd Reg Coeff 1058.007 0.621 0.322
(Weight)
Std beta 0.767 0.544 0.826
T / Critical Ratio 17.44 9.452 21.38
Mechanism of EFA:
Calculate Factor Coefficients and then Calculate Factor Scores:
F1 = a11Zs11 + a12Zs12 + a13Zs13+ a14Zs14+ …….
F1= (0.138 * 0.32846) + (0.14*0.85978) + …… = 1.24

F2 = a11Zs1 2+ a12Zs12 + a13Zs13+ a14Zs14+ …….

Mechanism of CFA:
Calculate Factor Score and then Calculate Factor Coefficients:
S1 = a1 F1 + e1
S2 = a2 F1 + e2
S3 = a3 F1 + e3
S4 = a4 F1 + e4
If e=0 then it’s a perfect prediction.
This will be checked by Chi square test.

Reporting CFA
Results

Regression Results Validity Issues Model fit Indices

Chi-square Goodness of fit Badness of fit

p<0.05 Content Validity Construct Validity
Statistics indicator indicator

1. Through ROL Convergent Discriminant CFI, GFI, IFI, NFI,

RMSEA, RMR etc
2. Expert Opinion Validity Validity AGFI etc

1. CR>0.7
1. AVE>MSV
2. AVE>0.5
2. AVE>ASV
3. CR>AVE
ORDERS OF CFA

Zero Order CFA

Second Order CFA

Third Order CFA
 First Order CFA
(of one second order and
one zero order)
SEM
Research Work:
Initial Model after Review of Literature

SPSS Data File: Delldata.sav

A = s1, s2, s3, s4
B = s5, s8, s9, s10
C = s11, s12, s13
Revised Model

Steps:
1. Check all three Constructs in First order CFA – Regression, Validity and Model fit results
2. Check for the Mediation effect (Through Imputation or Constructs)
Models in AMOS
 A -> B -> C Feedback not complete Recursive Model
 A -> B -> C ->A Feedback complete Non-Recursive Model

Improving Results
 Model Fit indices by Modification Indices

 Convergent Validity by Removing Outliers

CR AVE MSV ASV B A C
B 0.877 0.642 0.350 0.233 0.801    
A 0.905 0.707 0.350 0.306 0.592 0.841  
C 0.929 0.814 0.261 0.188 0.340 0.511 0.902

CR AVE MSV ASV B A C

B 0.889 0.668 0.420 0.266 0.817    
A 0.912 0.724 0.420 0.345 0.648 0.851  
C 0.953 0.872 0.270 0.191 0.334 0.520 0.934

 Discriminant Validity by Removing Cross-loading