Syntax - Vanderbilt Business School

Report
Donde Esta Lisrel? http://www.ssicentral.com/
be forewarned: Lisrel is not Mac-friendly
You can download a “free 15 day trial edition”
(the program has everything, but it goes “poof”
in 2 weeks), or you can download a “free
student edition” (which remains accessible
forever but has truncated options, e.g., fewer
variables are allowed, etc.):
Small Example to Play with…
Lisrel Syntax for Path Model
path example miniwoohoo.
da ni=5 no=100 ma=cm
cm sy
2.0113
-0.0719
2.5384
1.1926 -1.2155 2.0738
1.2482 -0.4925 0.8035
0.7615 -0.9282 0.3866
la
q c v cs r
se
v cs r q c /
Title end w period.
“da”=data, ni=#input vars,
no=sample size, ma=cm
says analyze the cov matr
2.0076
1.1610
“la” = labels,
next line list your vars (ni of them).
1.9957
“se” = select,
endogenous first, then exogenous.
mo ny=3 ne=3 nx=2 nk=2 lx=id,fi td=ze,fi ly=id,fi te=ze,fi ph=st,fr be=fu,fr ga=fu,fr
pa be
0 0 0
1 0 0
0 1 0
pa ga
1 1
1 0
0 1
pd
ou
“pa”=pattern, 0 means
set fixed (to zero), 1 means
estimate this parameter.
“mo”=model, nx=#xvars, nk=#ksi
constructs/factors, lx=factor ldgs
matrix, “fu,fr”=full, free, td=meas
error, “di,fr”=diag,free
“pd”= draw a path diagram for me.
“ou”= output
Lisrel Notation
 relates  ‘s to  ‘s
column goes to/ affects/causes row
 relates  ‘s to other  ‘s
 
column goes to row
 1  quality
 2  cos t
 1  value
 11
 12
 2  cs / d
 21
0
 3  repeat
0
 32
B 
1
2
3
 1  value
0.
0
0
 2  cs / d
 21
0.
0
 3  repeat
0
 32
0.
column goes to
(or “affects” or
“causes” row)
Lisrel Notation
PSI (prediction/modeling)
errors on eta’s:
η1 η2 η3
η1 ζ1 0 0
η2 0 ζ2 0
η3 0 0 ζ3
PHI (intercorrelations
among the
exogeneous
constructs, ksi’s)
ξ1 ξ2
ξ1 1 φ
ξ2 φ 1
Small Example to Play with…
Lisrel Syntax for Path Model
path example miniwoohoo.
da ni=5 no=100 ma=cm
cm sy
1.00000
-0.03183
1.00000
0.58395
-0.52975
0.62116
-0.21818
0.38010
-0.41242
la
q c v cs r
se
v cs r q c /
1.00000
0.39381
0.19005
1.00000
0.58002
1.00000
mo ny=3 ne=3 nx=2 nk=2 lx=id,fi td=ze,fi ly=id,fi te=ze,fi ph=st,fr be=fu,fr ga=fu,fr
pa be
0 0 0
1 0 0
0 1 0
pa ga
1 1
1 0
0 1
ou
Path Model Key Output
BETA
v
-------- -
cs
-------- -
r
-------- -
cs
0.05
(0.10)
0.49
- -
- -
r
- -
0.51
(0.08)
6.73
- -
q
-------0.57
(0.06)
8.95
c
--------0.51
(0.06)
-8.07
0.59
(0.10)
6.13
- -
- -
-0.30
(0.08)
-3.93
v
Parameter estimate
Standard error
t-statistic
GAMMA
v
cs
r
Goodness of Fit Statistics
Degrees of Freedom = 5
Minimum Fit Function Chi-Square = 29.77 (P = 0.00)
Comparative Fit Index (CFI) = 0.86
Standardized RMR = 0.073
The Lisrel Model
Measurement Model:
Structural Model:
y   y  
x   x  
  B    
Definitions:
y is a vector of observed indicators of the dependent latent endogenous variables
x is a vector of independent, or exogenous, variables
Λy is the factor loadings of y on η
Λx is the matrix of factor loadings of of x on ξ
η is a vector of latent dependent, or endogenous constructs
ξ is a vector of the independent latent variables, exogenous constructs
ε is vector of measurement errors in y
δ is vector of measurement errors in x
Γ is a matrix of coefficients of the ξ’s on the η’s (the structural relationships)
B is a matrix of coefficients of the η’s on η’s (the structural relationship)
ζ is a vector of equation errors (random disturbances) trying to predict the
endogenous guys η (the structural relationship errors)
Figure 3: Consumer Evaluation Relationships per Culture, per Marketplace
Latin America
Goods
Cost
Service
Prod
SalesRep
Northern Europe
Services
Cost
Valu
e
Qual
Serv
Continue
Easy
Product
SalesRep
Cost
Continue
Easy
Value
Serv
SalesRep
Qual
Cost
Value
Prod
Value
Serv
Qual
Easy
Continue
Prod
SalesRep
Qual
Continue
Easy
Iacobucci, Grisaffe, Duhachek and Marcati, “FAC-SEM: A Methodology for Modeling
Factorial Structural Equations Models, Applied to Cross-Cultural and Cross-Industry Drivers of Customer Evaluations,”
Journal of Service Research. Research also featured in, “Mapping the World of Customer Satisfaction,” Harvard
Business Review.

similar documents