Sunteți pe pagina 1din 67

Introduction to Structural Equation Modelling

Partial Least Squares Path Modelling


(PLSPM)
Joaqun Alds Manzano
Universitat de Valncia
*joaquin.aldas@uv.es

Partial Least Squares (PLS)

In 1960 Karl Jreskog proposed an algorithm to estimate covariance based


structural models: maximum likelihood algorithm (ML)
in 1970 ML algorithm was made operational in the first widely used
commercial software: LISREL
Herman Wold (Jreskog thesis supervisor) questioned the applicability of
covariance structure models estimation via ML because of its heavy
distribution assumptions and high sample sizes needed.
Wold proposed an alternative approach to these situations: Partial Least
Squares
The first general PLS algorithm Wold (1973) offered was called NIPALS
(=Non linear Iterative PArtial Least Squares) and a final presentation of the
PLS approach to path modelling with latent variables is present in Wolds
(1982) paper.

Partial Least Squares



Covariance based SEM rationale

X1

X2

X3

X1

1.00

X2

.087

1.00

X3

.140

.080

1.00

X4

.152

.143

.272

X4

1.00

Partial Least Squares



Covariance based SEM rationale
X1
.83

.33 .26

.46 .59

X2

X3

X1

1.00

X2

.087

1.00

X3

.140

.080

1.00

X4

.152

.143

.272

X4

1.00

0.33 0.26 = 0.085 ; 0.087 = X1 X 2

0.33 0.83 0.46 = 0.126 ; 0.140 = X1 X 3

Partial Least Squares



CBSEM solution vs. PLSPM solution

.83

.33 .26

.22

.46 .59

CBSEM solution

.75 .60

.54 .71

PLSPM solution

Partial Least Squares


Partial Least Squares



Terminology

Errors

Independent
Latent variable

Dependent
Latent variable
Independent
Latent variable

Manifest variables or indicators of


the independent construct

Manifest variables or indicators of


the dependent construct

Partial Least Squares



Terminology

Weights

Loadings

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Initialization

Inner weights estimation

Inside approximation

Outer weights estimation

Outside approximation

Stop criterion

Estimation of loadings,
weights and path coefficients

Raw
data

Manifest variable
indicators

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Initialization

is calculated
adding its indicators
(weights and
loadings = 1)

Step1

!1outer = x11 + x12


!2outer = x21 + x22
! outer = x + x + x
3

31

32

33

w11 = w12 = w21 = w31 = w32 = w33 = 1

10

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Inner weights estimation

Step 2

Covariance
between adjacent
LV is calculated

" cov(! outer , ! outer ) ! ,! adjacent


$
i
j
i
j
eij = #
$%
0
rest of cases

11

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Inner approximation

LV estimation
from their
previous outer
estimation

Step 3

!iinner = " eij! outer


j
j

#!1inner = e13!3outer
%% inner
outer
$!2 = e23!3
% inner
outer
outer
!
=
e
!
+
e
!
%& 3
13 1
23 2
12

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Outer weights estimation

Paso 4

Weights and
loading estimation

x jh = c jh + w jh! inner
+ " jh
j
# x31 = c31 + w31!3inner + " 31
%%
inner
$ x32 = c32 + w32!3 + " 32
%
inner
x
=
c
+
w
!
+ " 33
%& 33
33
33 3
' w 31 , w 32 , w 33

Reflective
13

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Outer weights estimation

Paso 4

Weights and
loading estimation

! inner
= c j + " w jh x jh + # j
j
h

$&!1inner = c1 + w11 x11 + w12 x12 + #1


% inner
&'!2 = c2 + w21 x21 + w22 x22 + # 2
( w 11 , w 12 , w 21 , w 22
Formative
14

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Outer approximation

Step 5

is calculated from
the last loadings &
weights estimation

outer
= w jh x jh
j
h

1outer = w11 x11 + w12 x12


outer
2 = w21 x21 + w22 x22
outer
3 = w31 x31 + w32 x32 + w33 x33
15

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Stop criterion

Step 6

"

We stop iterating
if stop criterion
holds

k
k !1
w(jh ) ! w(jh ) < 10!5

h, j

k
k !1
k
k !1
w11( ) ! w11( ) + w12( ) ! w12( ) +

(k )
( k !1)
(k )
( k !1)
w21
! w21
+ w22
! w22
+
(k )
( k !1)
(k )
( k !1)
(k )
( k !1)
w31
! w31
+ w32
! w32
+ w33
! w33
< 10!5
16

Partial Least Squares



PLSPM algorithm rationale
(Fornell, Barclay & Rhee, 1988; Wold, 1966)

Final estimation

Step 7

! j = ! outer
j

Final parameter
estimation is
performed

Latent variable scores

x jh = c jh + ! jh" j + # jh

" j = c j + % $ jh x jh +# j

reflective
Loadings
formative
Weights

! j = # " ij!i +$ j

Structural paths

17

Partial Least Squares


PLSPM specification restrictions

Structural part of the model must be recursive


Non-recursive paths are not allowed
No logical loops
Each latent variable must be connected at least to another LV (*)
Each LV must be measured by at least one indicator (second order factors?)
An indicator can be associated only to one LV
The model cannot be formed by non-connected blocks (*)

(*) SmartPLS restrictions, not necessarily PLSPM restrictions

18

Partial Least Squares


When using PLSPM instead of CBSEM?
Chin y Newsted (1999; p.337)

Small although representative sample sizes (to be clarified)


The phenomenon in question is relatively new or changing and
the theoretical model or measures are not well formed,
The data conditions relating to normal distribution,
independence, and/or sample size are not met
There is an epistemic need to model the relationship between
LVs and indicators using reflective measures
The objective is prediction

19

Partial Least Squares



What means a small sample size?
Minimum sample size is 10 times the higher of the
previous calculations:

5 10 = 50

Higher number or LV that point to a dependent


LV (3)

Number of indicators of the most complex


forrmative LV (5)

20

Partial Least Squares


What means a small sample size?

Chin & Newsted (1999; p.314); Hair, Anderson, Tatham & Black (1995; p.2 y p.10-13)

Previous criterion is the classical rule of thumb, but power analysis must be
considered
Test power:
Probability of rejecting the null hypothesis when it is false
(Type I error): probability of rejecting the null hypothesis when it is true
(Type II error): probability of not rejecting the null hypothesis when it is
false
1 (Power): probability of rejecting the null hypothesis when it is false
In Social Sciences, Minimum 80% (Cohen, 1988)
Does the previous rule of thumb meet this criterion?
For 5 predictors (our model), =.05 y and a moderate effect size, the power of
the test would be:
N
1
20

.16

40

.37

50

.47

60

.57

100

.84

L
21

Partial Least Squares


Recommendation for performing power analysis

Use a good power analysis software like G*Power


Can be downloaded for free at:
http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/
Read the paper:
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A
flexible statistical power analysis program for the social, behavioral, and
biomedical sciences. Behavior Research Methods, 39, 175-191.

22

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

23

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Data screening
Missing data imputation with adequate software, as listwise deletion is the
default for SmartPLS and only media imputation is available in actual beta.
Draw the model
Run the model
Very few options
Only weighting scheme is configurable (marginal effect on the result)
Validating the measurement model
LV Reliability (Cronbachs alpha)
LV Reliability (Composite reliability)
Convergent validity (loadings size and significance, AVE, cross-loadings)
Discriminant validity (AVE vs. LV correlation)
Structural model evaluation
Explained variance of dependent LV (R squared)
Predictive relevance (Q2 using blindfolding)
Structural paths significance (bootstrapping)
Global fit
No indices

24

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Data screening
Missing data imputation with adequate software, as listwise deletion is the
default for SmartPLS and only media imputation is available in actual beta.

Missing value coding schema must be indicated to the software. -1 coding used here

25

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Draw the model

26

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Run the model


Very few options
Only weighting scheme is configurable (marginal effect on the result)

27

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

Reliability (Cronbachs alpha)


Benchmark >.70 (Nunnally & Bernstein, 1994)
Composite reliability (Wets, Linn, Joreskog, 1974; Fornell & Larcker, 1981)
Only applied to reflective constructs, not to formative ones (Chin, 1998)
Benchmark > 0.6 (Bagozzi & Yi, 1988)
Calculated for each LV

CR =

L
ij

L
- ij + - Var Eij

j
j

( )

Where Lij is the standardized loading of item j on factor i


Var (Eij) is the error variance, calculated as:

( )

Var Eij = 1 ! Lij2


28

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model


ACTITUD
DEP1
DEP2
DEP3
DEP4
DEP5
DEP6
p.03.01
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.08
p.03.09
p.03.10
p.03.11
p.03.12
p.04.01
p.04.02
p.04.03
p.04.04
p.04.05
p.04.06
p.04.07
p.04.08
p.04.09
p.04.10
p.05

DEPEND
0,5274
0,3278
0,4300
0,9250
0,6135
0,5955

FACUSO

INTCOMP

UTIPER

Outer loadings

0,5714
0,7917
0,8203
0,8098
0,7766
0,7976
0,8198
0,4614
0,6897
0,7444
0,3946
0,7441

CR=.8938

0,6060
0,6617
0,2838
0,7607
0,8330
0,8042
0,7570
0,5889
0,4969
0,6737
1,0000

(0.5714 + 0.7917 + 0.8203 + 0.8098 + 0.7766 + 0.7976)


CR =
(0.5714 + 0.7917 + 0.8203 + 0.8098 + 0.7766 + 0.7976) + (1! 0.5714) + (1! 0.7917) + (1! 0.8203)
2

+ (1! 0.8098) 2 + (1! 0.7766) 2 + (1! 0.7976) 2

= 0.8938

29

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

SmartPLS automatically calculates CR and Cronbachs alphas:

AVE Composite Reliability R Square Cronbachs Alpha


0,4425
0,8823 0,4375
0,8518
0,4047
0,4374
0,8148
0,7616
1,0000
1,0000 0,3210
1,0000
0,5869
0,8938 0,2401
0,8557

Communality
0,4425
0,3595
0,4374
1,0000
0,5869

Redundancy
0,0708
0,0412
0,1940
0,1402

30

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

Reliability
Average Variance Extracted criterion (Fornell & Larcker, 1981):

!L

ij

AVEi =

!L

ij

( )

+ !Var Eij
j

Same notation CR
Benchmark > 0,5 (Fornell & Larcker, 1981)
Only for reflective LV, not formative ones (Chin, 1998)

31

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model


ACTITUD
DEP1
DEP2
DEP3
DEP4
DEP5
DEP6
p.03.01
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.08
p.03.09
p.03.10
p.03.11
p.03.12
p.04.01
p.04.02
p.04.03
p.04.04
p.04.05
p.04.06
p.04.07
p.04.08
p.04.09
p.04.10
p.05

AVE =

DEPEND
0,5274
0,3278
0,4300
0,9250
0,6135
0,5955

FACUSO

INTCOMP

UTIPER

Outer loadings

0,5714
0,7917
0,8203
0,8098
0,7766
0,7976
0,8198
0,4614
0,6897
0,7444
0,3946
0,7441

AVE=.5869

0,6060
0,6617
0,2838
0,7607
0,8330
0,8042
0,7570
0,5889
0,4969
0,6737
1,0000

0.57142 + 0.7917 2 + 0.82032 + 0.80982 + 0.77662 + 0.79762


= 0.5896
0.5714 + 0.7917 + 0.8203 + 0.8098 + 0.77662 + 0.79762 + (1 ! 0.5714)2 + (1 ! 0.7917)2 + (1 ! 0.8203)2 + (1 ! 0.8098)2 + (1 ! 0.7766)2 + (1 ! 0.7976)2
2

32

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

SmartPLS automatically calculates AVE

AVE Composite Reliability R Square Cronbachs Alpha


0,4425
0,8823 0,4375
0,8518
0,4047
0,4374
0,8148
0,7616
1,0000
1,0000 0,3210
1,0000
0,5869
0,8938 0,2401
0,8557

Communality
0,4425
0,3595
0,4374
1,0000
0,5869

Redundancy
0,0708
0,0412
0,1940
0,1402

33

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

Convergent validity (Same criteria than those from SEM)


Benchmark > 0.7 (Carmines & Zeller, 1979)
Some researchers believe this criterion is too strict when new scales are
being developed (Barclay, Higgins, Thompson, 1995; Chin, 1998)
Bagozzi & Yi (1988), for instance, propose > 0.6
Notice! Formative constructs are evaluated in terms of their weights
(Chin, 1998), focusing in their significance and not in their size.
Multicollinearity must be tested.
And what about significance? We will go back to it later.

34

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model


ACTITUD
DEP1
DEP2
DEP3
DEP4
DEP5
DEP6
p.03.01
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.08
p.03.09
p.03.10
p.03.11
p.03.12
p.04.01
p.04.02
p.04.03
p.04.04
p.04.05
p.04.06
p.04.07
p.04.08
p.04.09
p.04.10
p.05

DEPEND
0,5274
0,3278
0,4300
0,9250
0,6135
0,5955

FACUSO

INTCOMP

UTIPER

0,5714
0,7917
0,8203
0,8098
0,7766
0,7976

Formative constructs:
focus on WEIGHTS
not on loadings

UTIPER1 should be
dropt off

0,8198
0,4614
0,6897
0,7444
0,3946
0,7441
0,6060
0,6617
0,2838
0,7607
0,8330
0,8042
0,7570
0,5889
0,4969
0,6737
1,0000

Outer loadings

35

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

Convergent validity
In SEM Lagrange multiplier test allowed us to detect items with
significant loadings in more than one LV, what is a threat to convergent
validity
PLS offers cross-loadings of one indicator on all the LV. If the loading of
one item on a factor different to the one it was designed for is higher
than the loading on that factor, the item should be deleted.
Directly calculated by SmartPLS.
Only apply to reflective constructs.

36

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model


DEP1
DEP2
DEP3
DEP4
DEP5
DEP6
p.03.01
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.08
p.03.09
p.03.10
p.03.11
p.03.12
p.04.01
p.04.02
p.04.03
p.04.04
p.04.05
p.04.06
p.04.07
p.04.08
p.04.09
p.04.10
p.05

ACTITUD
0,2843
0,2736
0,3092
0,4620
0,3779
0,3635
0,4120
0,5003
0,4848
0,5004
0,4686
0,5295
0,3378
0,0874
0,2638
0,3360
0,0771
0,3617
0,6060
0,6617
0,2838
0,7607
0,8330
0,8042
0,7570
0,5889
0,4969
0,6737
0,4981

DEPEND
0,5274
0,3278
0,4300
0,9250
0,6135
0,5955
0,2931
0,5620
0,5152
0,4562
0,4622
0,4980
0,3596
0,0832
0,2390
0,3571
0,0938
0,3942
0,3289
0,4683
0,1295
0,3674
0,4246
0,4296
0,3454
0,2390
0,2470
0,3567
0,4096

FACUSO
0,2783
0,0460
0,1283
0,4160
0,2156
0,2098
0,2941
0,4143
0,4089
0,4371
0,3351
0,3454
0,8198
0,4614
0,6897
0,7444
0,3946
0,7441
0,1553
0,4597
0,0061
0,2600
0,3923
0,4008
0,1740
0,2127
0,0919
0,3685
0,4597

INTCOMP
0,2490
0,0335
0,0911
0,3823
0,1844
0,2053
0,3414
0,4306
0,3886
0,4497
0,3521
0,4032
0,3756
0,2349
0,3333
0,3577
0,1633
0,3155
0,2226
0,3545
-0,0157
0,3728
0,3995
0,4752
0,3224
0,3278
0,2122
0,3572
1,0000

UTIPER
0,2735
0,2142
0,2691
0,5852
0,3926
0,3636
0,5714
0,7917
0,8203
0,8098
0,7766
0,7976
0,3862
0,0687
0,3174
0,3976
0,0000
0,4214
0,3559
0,4331
0,1166
0,4526
0,5403
0,5991
0,3754
0,4138
0,3280
0,3683
0,5168

UTIPER

FACUSO

ACTITUD

37

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

Discriminant validity
Average variance extracted criterion (Fornell & Larcker, 1981)
One LV should share more variance with its indicators than with other LV
Squared correlation between two LV must be compared to the AVE of
each of the factors
Discriminant validity holds if:

AVEi > !ij2


AVE j > !ij2

Only applies to reflective constructs, not formative ones (Chin, 1998)

38

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Validating the measurement model

SmartPLS offers all the information we need: AVE for each latent variable and
the correlations among factors

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

AVE
0,4425
0,0000
0,4374
1,0000
0,5869

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

ACTITUD
1,0000
0,5224
0,4246
0,4981
0,6315

DEPEND

FACUSO INTCOMP

UTIPER

1,0000
0,4446
0,4096
0,6145

1,0000
0,4597
0,4900

1,0000

1,0000
0,5168

As a matter of example, for the two critical factors (higher correlation):


2
!utiper
= 0.63152 = 0.3988
" actitud

AVEutiper = 0.5869 > 0.3988


AVEactitud = 0.4425 > 0.3988
39

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Evaluating the structural model


Explained variance of the dependent LV (R squared)
Based on each regression R2
R2 are interpreted as in any multiple regression: the amount of variance of the
dependent LV explained by the set of independent LV
R2 should never be < 0.1 (Falk & Miller, 1992)
SmartPLS provides R2 both in the output and in the charts
My opinion: Falk & Miller (1992) is an arbitrary criterion, power analysis
should be added

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

AVE Composite Reliability R Square Cronbachs Alpha


0,4425
0,8823 0,4375
0,8518
0,4047
0,4374
0,8148
0,7616
1,0000
1,0000 0,3210
1,0000
0,5869
0,8938 0,2401
0,8557

Communality
0,4425
0,3595
0,4374
1,0000
0,5869

Redundancy
0,0708
0,0412
0,1940
0,1402

40

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

R2

41

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Evaluating the structural model


Predictive relevance (Q2 using blindfolding)
R2 size must be complemented (if not substituted) by Stone (1974) & Geisser
(1975) resampling proposal.
Blindfolding, implies deleting one part of the data of a dependent LV and
estimating them again using the structural part of the model. This process
repeats until each data point has been deleted and estimated.
Omission distance D has to be decided. It indicates the percentage of data
that are going to be deleted in each step. It must not be a perfect divisor of
the sample size. Wold (1982) recommends numbers between 5 and 10.
Let us assume that case n of variable k has been omitted. We can estimate it
by regressing that variable on the rest of the LV attending the model
structure. But as we know its true value, we can calculate the difference.
Adding those differences for all the omitted data for variable k we get:

Ek = " ykn ! y kn
n=1

42

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Evaluating the structural model


Predictive relevance (Q2 using blindfolding)
Now let us assume that, instead of estimating it by using the structural part of
the model, we estimate it ignoring the model by using variable k mean. Once
again we can calculate de difference and repeating it for all the omitted items:
N

Ok = " ykn ! ykn


n=1

Q2 statistic for each dependent LV can be calculated as follow:

Qk2 = 1 !

Ek
Ok

Q2 is a measure of how well the observed values can be replicated using the
estimated parameters:
Q2>0 implies the model has predictive relevance regarding variable k
Q2<0 implies the model lacks of predictive relevance regarding variable k

43

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Evaluating the structural model


Predictive relevance (Q2 using blindfolding)

SmartPLS directly provides Q2 values for all the dependent LV

Total
ACTITUD
DEPEND
INTCOMP
UTIPER

SSO
4650,0000
2790,0000
465,0000
2790,0000

SSE
3805,8981
2414,2583
326,3057
2414,4798

1-SSE/SSO
0,1815
0,1347
0,2983
0,1346

44

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Evaluating the structural model


Analysing the significance of structural paths (bootstrapping)
Evaluating the measurement model
Convergent validity: testing the significance of loadings (bootstrapping)
Bootstrapping
It is a resampling procedure through which N random samples are generated
from the original sample with replacement.
Average values of the parameters estimated in each sample are calculated
and compared to the parameters obtained in the original sample.
Big differences lead to no significant parameters.
Number of subsamples N must be high (around N=500)
Each subsample must have the same sample size that the original sample.

45

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping

One
ofdethe
subsamples
Una
la N N
submuestras

Muestra original
Original
sample

20 20 20 20 20 20 20

20

20 20 20 20 20 20 20
4

20

46

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping
DEP1 -> DEPEND
DEP2 -> DEPEND
DEP3 -> DEPEND
DEP4 -> DEPEND
DEP5 -> DEPEND
DEP6 -> DEPEND
p.03.01 <- UTIPER
p.03.02 <- UTIPER
p.03.03 <- UTIPER
p.03.04 <- UTIPER
p.03.05 <- UTIPER
p.03.06 <- UTIPER
p.03.07 <- FACUSO
p.03.08 <- FACUSO
p.03.09 <- FACUSO
p.03.10 <- FACUSO
p.03.11 <- FACUSO
p.03.12 <- FACUSO
p.04.01 <- ACTITUD
p.04.02 <- ACTITUD
p.04.03 <- ACTITUD
p.04.04 <- ACTITUD
p.04.05 <- ACTITUD
p.04.06 <- ACTITUD
p.04.07 <- ACTITUD
p.04.08 <- ACTITUD
p.04.09 <- ACTITUD
p.04.10 <- ACTITUD
p.05 <- INTCOMP

Original Sample
Standard T
Sample
Mean
Error
Statistics
0,5274
0,5192
0,0727
7,2547
0,3278
0,3213
0,0762
4,3029
0,43
0,423
0,0777
5,5345
0,925
0,9149
0,0266 34,7133
0,6135
0,6033
0,0673
9,121
0,5955
0,5856
0,0616
9,6686
0,5714
0,5668
0,0449 12,7149
0,7917
0,7911
0,0205 38,5845
0,8203
0,8193
0,019 43,2043
0,8098
0,8079
0,0209 38,8366
0,7766
0,7727
0,0256 30,3694
0,7976
0,7945
0,0223 35,7663
0,8198
0,8163
0,0204 40,2304
0,4614
0,4544
0,0808
5,7132
0,6897
0,685
0,0432 15,9552
0,7444
0,7425
0,03 24,8401
0,3946
0,3861
0,0907
4,3502
0,7441
0,7426
0,0286 25,9782
0,606
0,6014
0,0447 13,5555
0,6617
0,6619
0,0389 17,0137
0,2838
0,2813
0,0739
3,8391
0,7607
0,7576
0,0289 26,3512
0,833
0,8304
0,017 49,0118
0,8042
0,8032
0,0205 39,2634
0,757
0,7543
0,0299 25,2988
0,5889
0,5857
0,0417 14,1064
0,4969
0,4916
0,0464 10,7014
0,6737
0,6717
0,0384 17,5488
1
1
0
0

Outer loadings

47

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping

DEP1 -> DEPEND


DEP2 -> DEPEND
DEP3 -> DEPEND
DEP4 -> DEPEND
DEP5 -> DEPEND
DEP6 -> DEPEND
p.03.01 <- UTIPER
p.03.02 <- UTIPER
p.03.03 <- UTIPER
p.03.04 <- UTIPER
p.03.05 <- UTIPER
p.03.06 <- UTIPER
p.03.07 <- FACUSO
p.03.08 <- FACUSO
p.03.09 <- FACUSO
p.03.10 <- FACUSO
p.03.11 <- FACUSO
p.03.12 <- FACUSO
p.04.01 <- ACTITUD
p.04.02 <- ACTITUD
p.04.03 <- ACTITUD
p.04.04 <- ACTITUD
p.04.05 <- ACTITUD
p.04.06 <- ACTITUD
p.04.07 <- ACTITUD
p.04.08 <- ACTITUD
p.04.09 <- ACTITUD
p.04.10 <- ACTITUD
p.05 <- INTCOMP

Original Sample
Standard T
Sample
Mean
Error
Statistics
0,5274
0,5192
0,0727
7,2547
0,3278
0,3213
0,0762
4,3029
0,43
0,423
0,0777
5,5345
0,925
0,9149
0,0266 34,7133
0,6135
0,6033
0,0673
9,121
0,5955
0,5856
0,0616
9,6686
0,5714
0,5668
0,0449 12,7149
0,7917
0,7911
0,0205 38,5845
0,8203
0,8193
0,019 43,2043
0,8098
0,8079
0,0209 38,8366
0,7766
0,7727
0,0256 30,3694
0,7976
0,7945
0,0223 35,7663
0,8198
0,8163
0,0204 40,2304
0,4614
0,4544
0,0808
5,7132
0,6897
0,685
0,0432 15,9552
0,7444
0,7425
0,03 24,8401
0,3946
0,3861
0,0907
4,3502
0,7441
0,7426
0,0286 25,9782
0,606
0,6014
0,0447 13,5555
0,6617
0,6619
0,0389 17,0137
0,2838
0,2813
0,0739
3,8391
0,7607
0,7576
0,0289 26,3512
0,833
0,8304
0,017 49,0118
0,8042
0,8032
0,0205 39,2634
0,757
0,7543
0,0299 25,2988
0,5889
0,5857
0,0417 14,1064
0,4969
0,4916
0,0464 10,7014
0,6737
0,6717
0,0384 17,5488
1
1
0
0

Outer weights

48

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping
Path coefficients

ACTITUD -> INTCOMP


DEPEND -> ACTITUD
DEPEND -> INTCOMP
FACUSO -> ACTITUD
FACUSO -> DEPEND
FACUSO -> UTIPER
UTIPER -> ACTITUD
UTIPER -> DEPEND
UTIPER -> INTCOMP

Original

Sample

Sample (O)

Mean (M)

0,2655
0,1891
0,0906
0,1159
0,1888
0,49
0,4585
0,522
0,2935

Standard
0,2673
0,1944
0,0968
0,1214
0,1918
0,4961
0,4513
0,5234
0,2834

Error (STERR)

0,0591
0,0502
0,0593
0,0479
0,0524
0,0418
0,0537
0,044
0,0583

T Statistics
(|O/STERR|)

4,4887
3,7677
1,5277
2,4185
3,5989
11,7114
8,5314
11,8668
5,0363

49

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Measurement model evaluation


Indicator 1 of usefulness is deleted (loading <.70)
Indicators 2 and 5 of ease of use are deleted (loading <.70)
No reliability problems
Regarding convergent validity some problems arise with AVE, but we wait to
see what happens after deleting de above mentioned items
No problems of discriminant validity
Structural model evaluation
We must not do anything until the measurement model is considered the
definitive one.
Once modified the measurement model, the estimation is performed again

50

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

51

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

ACTITUD
DEP1
DEP3
DEP4
DEP5
DEP6
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.09
p.03.10
p.03.12
p.04.04
p.04.05
p.04.06
p.04.07
p.04.10
p.05

DEPEND
0,5275
0,4170
0,9233
0,6106
0,5986

FACUSO

INTCOMP

UTIPER

0,7947
0,8339
0,8122
0,7910
0,8119

Convergent validity
Outer loadings

0,8146
0,6910
0,7508
0,7547
0,7914
0,8645
0,8207
0,7614
0,7110
1,0000

52

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Reliability: CR, Cronbachs alpha,


AVE

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

AVE
0,6265

Composite Reliability
0,8931

0,5686
1,0000
0,6543

0,8401
1,0000
0,9044

R Square Cronbachs Alpha


0,3874
0,8506
0,4112
0,7480
0,3193
1,0000
0,2512
0,8679

Communality
0,6265
0,4072
0,5686
1,0000
0,6543

Redundancy
0,0834
0,0539
0,1995
0,1633

53

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Convergent validity: crossloadings

DEP1
DEP3
DEP4
DEP5
DEP6
p.03.02
p.03.03
p.03.04
p.03.05
p.03.06
p.03.07
p.03.09
p.03.10
p.03.12
p.04.04
p.04.05
p.04.06
p.04.07
p.04.10
p.05

ACTITUD
0,2784
0,2810
0,4289
0,3501
0,3444
0,4843
0,4630
0,4795
0,4553
0,5066
0,3237
0,2492
0,3246
0,3640
0,7914
0,8645
0,8207
0,7614
0,7110
0,4942

DEPEND
0,5275
0,4170
0,9233
0,6106
0,5986
0,5633
0,5155
0,4548
0,4593
0,4981
0,3624
0,2397
0,3599
0,3946
0,3652
0,4226
0,4288
0,3414
0,3564
0,4126

FACUSO
0,2772
0,1448
0,4255
0,2365
0,2264
0,4273
0,4253
0,4537
0,3552
0,3582
0,8146
0,6910
0,7508
0,7547
0,2704
0,4021
0,4080
0,1866
0,3663
0,4566

INTCOMP
0,2490
0,0911
0,3823
0,1844
0,2053
0,4306
0,3886
0,4497
0,3521
0,4032
0,3756
0,3333
0,3577
0,3155
0,3728
0,3995
0,4752
0,3224
0,3572
1,0000

UTIPER
0,2676
0,2670
0,5889
0,3986
0,3724
0,7947
0,8339
0,8122
0,7910
0,8119
0,3733
0,3064
0,3966
0,4182
0,4330
0,5312
0,5901
0,3648
0,3629
0,5023

UTIPER

FACUSO
ACTITUD

54

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

AVE
0,6265

Composite Reliability
0,8931

0,5686
1,0000
0,6543

0,8401
1,0000
0,9044

ACTITUD
FACUSO
INTCOMP
UTIPER

ACTITUD
1,0000
0,4242
0,4942
0,5911

R Square Cronbachs Alpha


0,3874
0,8506
0,4112
0,7480
0,3193
1,0000
0,2512
0,8679

FACUSO

INTCOMP

UTIPER

1,0000
0,4566
0,5012

1,0000
0,5023

1,0000

LV
Discriminant validity
AVE vs. correlations

Communality
0,6265
0,4072
0,5686
1,0000
0,6543

Redundancy
0,0834
0,0539
0,1995
0,1633

AVE

ACTITUD

.79

FACUSO

.75

INTCOMP

N/A

UTIPER

.81

55

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

ACTITUD
DEPEND
FACUSO
INTCOMP
UTIPER

AVE
0,6265

Composite Reliability
0,8931

0,5686
1,0000
0,6543

0,8401
1,0000
0,9044

R Square Cronbachs Alpha


0,3874
0,8506
0,4112
0,7480
0,3193
1,0000
0,2512
0,8679

Communality
0,6265
0,4072
0,5686
1,0000
0,6543

Redundancy
0,0834
0,0539
0,1995
0,1633

Structural model
R2

Fuente: Hair, Black, Babin, Anderson y Tatham (2006; p.195)

56

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Structural model
Q2, predictive relevance

Total
ACTITUD
DEPEND
INTCOMP
UTIPER

SSO
2325,0000
2325,0000
465,0000
2325,0000

SSE
1799,2618
1947,7690
323,8041
1981,4057

1-SSE/SSO
0,2261
0,1622
0,3036
0,1478

57

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping
Outer loadings

58

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Bootstrapping

Outer weights

59

Partial Least Squares


Desarrollo de un caso de ejemplo
Sanz, Ruiz, Alds (2008)

Bootstrapping
Path coefficients

60

Partial Least Squares


An annotated example
Sanz, Ruiz, Alds (2008)

Results in an academic paper


Measurement model: reliability and convergent validity
Factor

Indicator

F1. Usefulness

UTPER2
UTPER3
UTPER4
UTPER5
UTPER6
FACUSO1
FACUSO3
FACUSO4
FACUSO6
DIM1
DIM3
DIM4
DIM5
DIM6
ACT4
ACT5
ACT6
ACT7
ACT10

F2. Ease of use

F3. Internet dependence

F4. Attitude

Loading

Weight

0,795***
0,834***
0,812***
0,791***
0,812***
0,815***
0,691***
0,751***
0,755***
0,286***
-0,122*
0,732***
0,168**
0,203**
0,791***
0,864***
0,821***
0,861***
0,711***

t value
42,02
53,33
44,17
31,73
44,49
41,27
17,09
27,26
29,95
4,57
1,64
13,71
1,96
2,47
32,23
62,43
44,86
27,63
20,30

CA

CR

AVE

0,868

0,904

0,654

0,748

0,840

0,569

N/A

N/A

N/A

0,851

0,893

0,627

***p<.01; **p<.05; *p<.10; N/A = Do not apply

61

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Results in an academic paper

Measurement model: discriminant validity


F1
F1
F2
F3
F4

F2
.81
0,501
0,618
0,591

F3
.75
0,459
0,424

F4
N/A
0,488

.79

**p<.01; *p<.05; N/A = Do not apply


Below diagonal: correlation among latent variables
Diagonal: square root of AVE

62

Partial Least Squares



An annotated example
Sanz, Ruiz, Alds (2008)

Results in an academic paper

Hypotheses testing
Hypothesis

Standardized
beta

H1: Ease of use Usefulness


H2: Ease of use Attitude
H3: Usefulness Attitude
H4: Usefulness Intention to buy
H5: Attitude Intention to buy
H6: Usefulness Internet dependency
H7: Ease of use Internet dependency
H8: Internet dependency Attitude
H9: Internet dependency Intention to buy

0,501**
0,138**
0,419**
0,268**
0,283**
0,518**
0,199**
0,166**
0,110*

t value
Bootstrap
13,38
3,14
8,34
4,49
5,05
13,54
4,04
3,48
2,08

R2 (Usefulness) = 0,251; R2 (Attitude) = 0,387; R2(Dependency) = 0,411; R2 (Intention) = 0,319;


Q2 (Usefulness) = 0,148; Q2 (Attitude) = 0,226; Q2(Dependency) = 0,162; Q2 (Intention) = 0,304;
**p<.01; *p<.05

63

Introduction to Structural Equation Modelling


Partial Least Squares Path Modeling


(PLSPM)
Exercise

Joaqun Alds Manzano


Universitat de Valncia
* joaquin.aldas@uv.es

64

Partial Least Squares


Determinants and consequences of client satisfaction
Fornell, Johnson, Anderson, Cha y Bryant (1996)

Model
Read paper from Fornell, Johnson, Anderson, Cha & Bryant (1996)

65

Partial Least Squares


Determinants and consequences of client satisfaction
Fornell, Johnson, Anderson, Cha y Bryant (1996)

Questions
Use data file ecsi.csv
Draw model of next figure
Evaluate if sample size is big enough to apply PLSPM
Evaluate measurement model reliability and validity and estimate
parameters significance using bootstrapping
Evaluate the structural model
Create a table to publish the results in an academic paper

66

Partial Least Squares



Determinants and consequences of client satisfaction
Fornell, Johnson, Anderson, Cha y Bryant (1996)

67

S-ar putea să vă placă și