2. Introduction
Baron and Kenny Mediation
Baron and Kenny Moderation
Andrew and Hayes Mediation
Andrew and Hayes Moderation
Conclusion
Activity
2
4. Referential papers
1.The Moderator-Mediator Variable Distinction in
Social Psychological Research: Conceptual,
Strategic, and Statistical Considerations
ABSTRACT
In 1986 Baron and Kenny set out to clarify the
differences between the terms “Moderation” and
“Mediation” as used in the social sciences.
2. Beyond Baron and Kenny: Statistical Mediation
Analysis in the New Millennium Andrew F. Hayes
5. Mediators
In an intervening variable model, variable X is
postulated to exert an effect on an outcome
variable Y through one or more intervening
variables, sometimes called mediators.
E.g: anxiety is acting as a mediator between high
performance work and counterproductive work
behavior.
9. What Is Moderation
The causal relationship from a causal variable or X to an
outcome or Y changes as a function of a moderator or M.
X and M interact to cause Y.
Effect of stress on mood is moderated by gender.
(Reuben M. Baron and David A. Kenny,1986)
10. Properties of Moderation
1.Desirable moderation:
zero-order correlation
2.Acts like a causal variable
3.Direction of the correlation changes.
(Reuben M. Baron and David A.
Kenny,1986)
12. EFFECT OF RELATION
The effect of X on Y changes by a
constant amount as M increases or
decreases as shown in the graph.
13. Statistical Estimation
Typically estimated as the interaction between X and M
Y = aX + bM + cXM + E
a = “main effect” of X
b = “main effect” of M
c = interaction between X and M
Important to include both X and M in the model.
14. Barron & Kenny (1986) mediation analysis based on the four steps
Independent
variable
Dependent
variable
Dependent
variable
Mediator
MediatorIndependent
variable
MediatorIndependent
variable
Dependent
variable
STEP 1
STEP 2
STEP 3
STEP 4
15. 15
Examine the intervening effect of anxiety on High performance wok system
and counterproductive work behavior.
WHATS?
IV…….
DV……
MEDIATOR….
29. Is it a Partial
mediation
or a Full
mediation?
30. Steps Measurement Standardized
Coefficients
β
t P
1 HPWS--CWB .567 12.5 0.000
2 HPWS--Anx .710 18.3 0.000
3 ANX--CWB .404 6.67 0.000
4 HPWS CWB .280 4.62 0.000
ANX .404 6.67 0.000
Table 1. Regression Analysis with High performance system as independent variable
Note: ∆R2=(0.08), F= (332,1)=110.3,p=0.000
31. 31
INTERPRETATION:
In order to examine the mediating impact of anxiety, all the four assumptions of Baron
and Kenny (1986) are used. Results demonstrate all the first three steps are significant.
In 4th step, high performance work system and anxiety are simultaneously regressed
on counterproductive work behavior. According to table in step 4 both p-values are
significant which is showing that anxiety has partial mediation effect on the
relationship of HPWS and CWB.
Further Table 1 represents that R2=.402 which tells that 40% variation in
counterproductive work behavior is due to the mediating effect. It can be seen that
there is an increase in R2 value of Stage 1 i.e. (.321) to R2 value of Stage 4 i.e. (.402).
Also there is a reduction in β values of both stages (.567,.404 ). With these and the
significant p-values at both stages show that there is a mediating effect of anxiety
between HPWS and CWB. It shows there is a partial mediation.
32. An alternative is to estimate the indirect effect and its significance using the Sobel
test (Sobel. 1982).
To test whether a mediator carries the influence of an IV to a DV.
The Sobel test works well only in large samples.
z-value = a*b/SQRT(b2*sa
2 + a2*sb
2)
a = B value (slope) for a-path
b = B value (slope) for b-path
sa = SE for a-path
sa = SE for b-path
Online Calculator for Sobel Test:
http://quantpsy.org/sobel/sobel.htm
Also available in the PROCESS macro discussed later
39. Path Direct Effect a Total Effect b Indirect
Effect
95% CI
Β P Β P Β Lower level Upp
er
leve
l
HPWS→
ANX→C
WB
.517(.048
)
.000 .394
(.032)
.000 -
.122(.037)
- 0.199 -
0.04
9
Note: →R2 = 0.567 ; F= 156.3; p=.000 * <0.05,
** p<0.01
Bootstrap standard error (shown in parenthesis)
a HPWS→CWB
b (HPWS→ANX) X(ANX→CWB)
40. .
Path Direct Effect a Total Effect b Indirect
Effect
95% CI
Β P Β P Β Lower level Upper
level
HPWS→ANX
→CWB
.195(.049) .000 .394
(.039)
.000 .199(.035) .1361 .279
Note: →R2 = 0.567 ; F= 156.3; p=.000 * <0.05, ** p<0.01
Bootstrap standard error (shown in parenthesis)
a HPWS→CWB
b (HPWS→ANX) X(ANX→CWB)
41. .
The results of macro are based on the re-sampling through bootstrapping. According to the
results exhibited in table show that all of the effects are significant i.e. total effect of HPWS
on CWB (X on Y) as (β= .394, P= .000) .This shows that anxiety has significantly mediating
role between the relationship of HPWS and CWB.
According to Preacher & Hayes (2008), in bootstrapped results, the Bias Corrected
Confidence Interval has two values (lower level and upper level). If zero exists between the
lower level and upper level values, then the variable will be fully mediating the relationship.
According to table it is shown that zero does not exist between the upper level and lower level
(lower level= -0.199, upper level= -0.049). This shows that there is partial mediation of
anxiety in the relationship between HPWS and CWB
INTEPRETATION
53. 53
B
SE β t Sig R2 ΔR2
F
Step1
High Performance
work system
0.340 .112 0.489 3.03 .003 .321 .321 156.3
Step2
Organization
Injustice
-.316 .111 -.526 -2.85 0.005 .586 .022 10.93
Step 3
High Performance
work system ×
Organization
Injustice
.055 .032 .523 1.736 0.084 .349 .006 3.012
Note: *p< .05, **p<.01
54. 54
In order to examine the moderating impact of organization injustice, all the three assumptions of Baron
and Kenny (1986) are used. Table shows the result of moderation effect of OIJ between HPWS and CWB. It
shows that the impact of both HPWS (β= .489, t=3.03, p=0.003) and OIJ (β= -.526, t= -2.85, p=0.005 ) are
significant. The overall model is also significant (0.000). In 3rd step after regressing the interaction term (i.e.
IV*MV), it didn’t produce a significant result (β=-.0523, t=1.74, p= .0.08). So there is no moderation.
59. .Table 15. Regression results for testing moderation of MOV Between B
and OCB
β SE t P R2 ΔR2 F UL
LL
Step 1
HPWS 0.339 .132 2.58 .01 .349 49.1
OIJ -0.316 .131 -2.41 .01
Step 2
HPWS ×
OIJ .055 .036 1.505 0.13 .349 49.36 -0.017
0.127
60. .
Table shows the result of moderation effect of OIJ between HPWS and CWB. It shows that in step 1, the
impact of both HPWS (β= .339, t=.112, p=0.003) and OIJ (β=-.316, t=-2.85, p=0.004) are significant.
The overall model is also significant (0.000) but After regressing the interaction term, it didn’t produce a
significant result (β=-.0549, t=-1.74, p= .0.08). If Zero exists between the upper level and lower level
values, then the variable doesn’t moderate the relationship. The result represented by table shows that
Zero exist between the upper level and lower level (Upper Level = .117, Lower Level =-.007). This shows
that OIJ doesn’t moderate the relationship between HPWS and CWB.
INTERPRETATION
66. 66
1.What is ideal type of moderation?
2.Indirect effect is product of which two pathways?
3.Which model is used in mediaton?
4.In which model do we compute interaction term by ourself?