11/10/2018

Assignment 6 and 7
• After controlling for the influence of income, is there
still a significant strong and positive relationship
between age and support for market economy?
• HA: After controlling for the influence of income, there
is still a significant strong and positive relationship
between age and support for market economy.
• HO: After controlling for the influence of income, there
is no longer a significant strong and positive
relationship between age and support for market
economy.

Correlations
Control Variables Age Support Income
-none-a Age Correlation 1.000 .834 .761
Significance (2-tailed) . .000 .000
df 0 16 16
Support Correlation .834 1.000 .637
Significance (2-tailed) .000 . .004
df 16 0 16
Income Correlation .761 .637 1.000
Significance (2-tailed) .000 .004 .
df 16 16 0

Income Age Correlation 1.000 .698

Significance (2-tailed) . .002
df 0 15
Support Correlation .698 1.000
Significance (2-tailed) .002 .
df 15 0
a. Cells contain zero-order (Pearson) correlations.

1
 11/10/2018

• Age is strongly and positively correlated with

support for the market economy and this is
very significant (r=.834, df=16, p<.001). Older
individuals tend to show greater support for
the market economy. On the other hand, age
has a very significant moderately high and
positive correlation with income (r=.761,
df=16, p<.001). Thus, older individuals are
likely to be earning higher incomes

Income and support for the market economy are

moderately and positively correlated and this is
very significant (r=.637, df=16, p<.001). As the
income of individuals increase, they are more likely
to support market economy. However, when the
partial correlation coefficient for age and support
for the market economy is computed, controlling
the effects of income, the level of correlation
slightly decreased (r=.698, df=15, p<.01). This
means that controlling for income had very little
effect on the strength of the relationship between
age and support for the market economy.

2
 11/10/2018

Assignment 2:
Using the Survey Data, conduct a Standard MR, Hierarchical MR
and Stepwise MR with Life Satisfaction (tlifesat) as the criterion
variable, and Optimism (toptim), Mastery (tmast) and Self-Esteem
(tslfest) as predictors

1. What is the R-Square of all the predictors combined? Write the

complete interpretation.
2. Based on the Hierarchical MR, which is/are best predictor/s of
Life Satisfaction?
3. Write the regression equation using the raw scores
(unstandardized) based on the outputs of the Hierachical MR.
4. Write the regression equation using the z-scores (standardized)
based on the output of the Stepwise MR. Provide the complete
interpretation.
*Paste all necessary SPSS tables per number

1. What is the R-Square of all the predictors

combined? Write the complete interpretation.

• Multiple regression analysis results show that

Optimism, Mastery and Self-Esteem
accounted for 32% of the variance in the
criterion variable, Life Satisfaction (R2=.320,
F(3,432)=67.209, p< .001).

3
 11/10/2018

2. Based on the Hierarchical MR, which is/are best

predictor/s of Life Satisfaction?

• An investigation of the individual contributions of each of

the three predictors namely, Self-Esteem, Optimism, and
Mastery, it was found that Self-Esteem had the biggest
contribution in the prediction of Life Satisfaction (β=.256,
p< .001). Optimism likewise significantly contributed to the
prediction of Life Satisfaction (β=.247, p< .001). The third
predictor, Mastery also showed significant predictive power
on Life Satisfaction (β=.170, p<.01). It can be said therefore
that the three predictors all significantly contribute to the
variance of Life Satisfaction.

3. Write the regression equation using the raw

scores (unstandardized) based on the outputs
of the Hierachical MR.
•
• Life Satisfaction = -3.137 + .321(Self-Esteem) +
.377(Optimism) + .294 (Mastery)

4
 11/10/2018

4. Write the regression equation using the z-scores (standardized) based on

the output of the Stepwise MR. Provide the complete interpretation.

• Life Satisfaction = .256(Self Esteem) + .247(Optimism) + .170 (Mastery)

• Stepwise multiple regression analysis was conducted to determine if Life

Satisfaction is predicted by Optimism, Mastery and Self-Esteem. There
were three emerging models from this analysis. The first model included
only Self- Esteem as predictor and this accounted for 23.6% of the
variance in the criterion variable, Life Satisfaction (R2=.236,
F(1,432)=133.373, p< .001). The second model included Self-Esteem and
Optimism as predictors and they determined 30.1% of the variance in Life
Satisfaction (R2=.301, F(2,432)=92.776, p< .001). The third model included
all three predictors and these variables explained 32% of the variance in
Life Satisfaction (R2=.320, F(3,432)=67.209, p< .001). All three variables
can therefore be used as predictors of Life Satisfaction.

MEDIATION AND MODERATION

ANALYIS

5
 11/10/2018

SIMPLE MEDIATION ANALYSIS

Mediation Analysis
• Examines how two variables precede or cause
a third variable
• Mediation implies a situation where the effect
of the IV on the DV can best be explained
using a third, “Mediator", variable.
• Mediation can be thought of as full / whole or
partial

6
 11/10/2018

Mediation Analysis

X Y
c
Total Effect

M
a b
Indirect Effect

c’
X Y
c
Direct Effect

Mediation Analysis
• Partial Mediation = 0 < c’ < c
• Complete Mediation = c’ = 0
• Total Effect = direct effect + indirect effect
c = c’ + ab
• Indirect Effect = amount of mediation;
reduction of the effect of the causal variable
(X) on the outcome (Y)
ab = c – c’

7
 11/10/2018

Mediation Analysis
• When to use:
1. Mediator provides a logical explanation for the
relation between IV and DV
2. Must be significant and substantial correlations
between
a. Predictor and Mediator variable
b. Mediator and Criterion variable
*correlation between Predictor and Criterion
doesn’t have to be significant or substantial

Mediation Analysis
• General Steps
1. Confirm the significance of the relationship between
the initial IV and MV (X M)
2. Confirm the significance of the relationship between
the initial IV and DV (X Y)
3. Confirm the significance of the relationship between
the initial MV and the DV in the presence of the IV
(M/X Y)
4. Confirm the insignificance (or the meaningful
reduction in effect) of the relationship between the
initial IV and the DV in the presence of the mediator
(X/M Y)

8
 11/10/2018

Mediation Analysis
• SPSS Steps – (Baron & Kenny, 1986)
Conduct Simple Regression between X
and M – determine significance

Conduct Simple Regression between M

and Y – determine significance

Conduct Simple Regression between X

and Y – determine significance

Conduct Multiple Regression Analysis

with X and M variables entered as “IV”

Interpret “Coefficients” Table (see next

slide)

Mediation Analysis
• Example
Use survey.sav
DV: Life Satisfaction
IV: Self-Esteem
MV: Optimism

9
 11/10/2018

Mediation Analysis
Step 1: Simple Regression between X and M
Model Summary

Std. Error of the

Model R R Square Adjusted R Square Estimate

1 .565a .319 .318 3.666

a. Predictors: (Constant), tslfest

Coefficientsa

Unstandardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) 6.533 1.111 5.881 .000

tslfest .465 .033 .565 14.214 .000

a. Dependent Variable: toptim

Mediation Analysis
Step 2: Simple Regression between M and Y
Model Summary

Adjusted R Std. Error of

Model R R Square Square the Estimate
1 .483a .233 .232 5.925
a. Predictors: (Constant), toptim

Coefficientsa

Unstandardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) 6.096 1.448 4.209 .000

toptim .737 .064 .483 11.482 .000

a. Dependent Variable: tlifesat

10
 11/10/2018

Mediation Analysis
Step 3: Simple Regression between X and Y
Model Summary

Adjusted R Std. Error of

Model R R Square Square the Estimate
1 .488a .238 .237 5.919
a. Predictors: (Constant), tslfest

Coefficientsa

Unstandardized Standardized
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) 1.809 1.790 1.010 .313

tslfest .613 .053 .488 11.629 .000

a. Dependent Variable: tlifesat

Mediation Analysis
Step 4: Multiple Regression of M and X on Y
Variables Entered/Removedb

Model Variables Entered Variables Removed Method

1 toptim, tslfest . Enter
a. All requested variables entered.
b. Dependent Variable: tlifesat

Model Summary

Change Statistics
Std. Error
R Adjusted of the R Square F Sig. F
Model R Square R Square Estimate Change Change df1 df2 Change
1 .549a .301 .298 5.667 .301 92.776 2 430 .000
a. Predictors: (Constant), toptim, tslfest

11
 11/10/2018

Mediation Analysis
Step 4: Multiple Regression of M and X on Y
Coefficientsa

Standar
dized
Unstandardized Coefficie Collinearity
Coefficients nts Correlations Statistics

Std. Zero- Parti Tolera

Model B Error Beta t Sig. order al Part nce VIF
1 (Constant) -1.146 1.785 -.642 .521

tslfest .390 .061 .311 6.375 .000 .486 .294 .257 .681 1.469

toptim .471 .074 .309 6.331 .000 .485 .292 .255 .681 1.469

a. Dependent Variable: tlifesat

Mediation Analysis
Optimism

a b
B =.465** B=.737*
*
c
B=.613** Life
Self-Esteem
Satisfaction

C’
B=.390**

12
 11/10/2018

Mediation Analysis
• Sample Interpretation
Regression analysis was used to investigate the hypothesis that Optimism
mediates the effect of Self-Esteem on Life Satisfaction. Results indicated
that Self-Esteem was a significant predictor of Optimism, B = .465, t =
14.214, p < .01, and that Optimism was a significant predictor of Life
Satisfaction, B = .737, t = 11.482 p < .01. These results support the
mediational hypothesis. Self-Esteem was still a significant predictor of Life
Satisfaction after controlling for the mediator, Optimism, B = .390, t =
6.375, p < .01. Its predictive power was slightly lower though than when
Optimism was factored into the relationship B = .613, t = 11.549, p < .01,
which is consistent with partial mediation. Approximately 30% of the
variance in Life Satisfaction was accounted for by the predictors (R2 = .301).

Mediation Analysis
• Using Hayes PROCESS

Choose
Choose Mediator
Analyze Independent
Variable (M)
Variable (X)

Choose Model
Choose Outcome
Regression Number (4 – for
Variable (Y)
simple mediation)

Linear PROCESS “OK”

13
 11/10/2018

Mediation Analysis
Run MATRIX procedure:
************* PROCESS Procedure for SPSS Release 2.16.1 *****************
Written by Andrew F. Hayes, Ph.D. www.afhayes.com
Documentation available in Hayes (2013). www.guilford.com/p/hayes3
*************************************************************************
*Model = 4
Y = tlifesat
X = tslfest
M = toptim
Sample size 433

*************************************************************************
Outcome: toptim
Model Summary
R R-sq MSE F df1 df2 p
.565 .319 13.439 202.045 1.000 431.000 .000
Model
coeff se t p LLCI ULCI
constant 6.533 1.111 5.881 .000 4.350 8.717
tslfest .465 .033 14.214 .000 .400 .529

Mediation Analysis
Outcome: tlifesat
Model Summary
R R-sq MSE F df1 df2 p
.549 .301 32.115 92.776 2.000 430.000 .000
Model
coeff se t p LLCI ULCI
constant -1.146 1.785 -.642 .521 -4.654 2.363
toptim .471 .074 6.331 .000 .325 .618
tslfest .390 .061 6.375 .000 .270 .511

************************** TOTAL EFFECT MODEL ***************************

Outcome: tlifesat
Model Summary
R R-sq MSE F df1 df2 p
.486 .236 35.028 133.373 1.000 431.000 .000
Model
coeff se t p LLCI ULCI
constant 1.934 1.794 1.079 .281 -1.591 5.460
tslfest .609 .053 11.549 .000 .506 .713

14
 11/10/2018

***************** TOTAL, DIRECT, AND INDIRECT EFFECTS ********************

Total effect of X on Y
Effect SE t p LLCI ULCI
.609 .053 11.549 .000 .506 .713
Direct effect of X on Y
Effect SE t p LLCI ULCI

Mediation Analysis
.390 .061 6.375 .000 .270 .511
Indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
toptim .219 .041 .145 .304
Partially standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
toptim .032 .006 .021 .045
Completely standardized indirect effect of X on Y
Effect Boot SE BootLLCI BootULCI
toptim .175 .032 .116 .241
Ratio of indirect to total effect of X on Y
Effect Boot SE BootLLCI BootULCI
toptim .359 .073 .231 .519
Ratio of indirect to direct effect of X on Y
Effect Boot SE BootLLCI BootULCI
toptim .561 .196 .300 1.078
R-squared mediation effect size (R-sq_med)
Effect Boot SE BootLLCI BootULCI
toptim .170 .028 .118 .230
Normal theory tests for indirect effect
Effect se Z p
.219 .038 5.772 .000

Mediation Analysis
• Sample Interpretation
In step 1 of the mediation model, the regression of Self-Esteem on
Optimism was significant (B = .465, t = 14.214, p < .01). Step 2 also
showed the mediator, Optimism (B = .737, t = 11.482 p < .01)
significantly predicted Life Satisfaction. These results support the
mediational hypothesis. Step 3 of the analysis revealed that the
predictor, Self-Esteem alone significantly predicts the criterion, Life
Satisfaction (B = .613, t = 11.549, p < .01). However, in step 4 of the
process, the predictive power of the predictor dropped slightly
lower when the mediatior, Optimism was controlled for (B = .613, t =
11.549, p < .01), consistent with partial mediation. A Sobel test was
conducted and found the partial mediation of Optimism on the
relationship between Self Esteem and Life Satisfaction significant
(z=5.772, p<.001).

15
 11/10/2018

MODERATOR ANALYSIS

MODERATOR ANALYSIS
• A moderator variable is one where there is a
different relationship between the main
variables in a study at different levels of the
moderator variable
• Vs Mediator: Mediator brings about the
relationship between two main variables in an
analysis; Moderator brings changes to the
magnitude of relationship between IV and DV

16
 11/10/2018

MODERATOR ANALYSIS
• Moderation implies an interaction effect
between the moderator and the predictor
variables.
Z

X Y

MODERATOR ANALYSIS
• SPSS Steps (thru Hierarchical Regression)
Standardize Raw Scores (convert to Z-
scores) [Analyze – Descriptive Statistics
–Descriptives – Save standardized
variables)

Create Interaction Terms of Z-scores

using Compute Function (IV *
Moderator)
Conduct Hierarchical Regression
Analysis
a. IV (first to enter)
b. Mod Variable (second to enter)
c. Interaction Term (last to enter)

17
 11/10/2018

MODERATOR ANALYSIS
• SPSS Steps (continued)
Check Model Summary.
If last Model is
significant, go on to
next step. Otherwise,
process stops.
Copy unstandardized
coefficients of last Model (under
coefficients table) to a new SPSS
data window. Label variables as
“a” for Constant, “b1” for IV,
“b2” for Moderator, “b3” for
Interaction Term

In the same window, create

Compute the predicted
new variables “X”, “M” and
criterion values “Y”.
“Y”. Enter the 3 values of
Transform – Compute
one standard deviation
Variable – Use the regression
above (1), below (-1) and the
equation with computed
at the mean (0) for X for all
“a”,”b1”, “b2”, and “b3”
values of M

MODERATOR ANALYSIS
• SPSS Steps (continued)
Plot the predicted
criterion values. Select “Y” for Dependent
Analyze – General Variable. Choose “X” and “M”
Linear Model - for the Fixed Factors.
Univariate

Click Plots. Select “X” for

Horizontal Axis, and “M” for
Edit the graph
Separate Lines. Click Add as needed

18
 11/10/2018

MODERATOR ANALYSIS
Use McNulty et al. (2008).sav
(McNulty et al. (2008) found a relationship between a
person’s Attractiveness and how much Support they
give their partner in newlyweds. Is this relationship
moderated by gender (i.e., whether the data were from
the husband or wife)?

DV: Support
IV: Attractiveness
Moderator: Gender

MODERATOR ANALYSIS
Model Summary
Change Statistics

R Adjusted R Std. Error of R Square F Sig. F

Model R Square Square the Estimate Change Change df1 df2 Change
1 .029 a .001 -.005 1.002664 .001 .135 1 162 .714
2 .064b .004 -.008 1.004127 .003 .528 1 161 .468
3 .292c .085 .068 .965310 .081 14.209 1 160 .000
a. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10)
b. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10), Gender
c. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10), Gender, Interaction Variable

ANOVAd
Model Sum of Squares df Mean Square F Sig.
1 Regression .136 1 .136 .135 .714a
Residual 162.864 162 1.005
Total 163.000 163
2 Regression .668 2 .334 .331 .718b
Residual 162.332 161 1.008
Total 163.000 163
3 Regression 13.908 3 4.636 4.975 .003c
Residual 149.092 160 .932
Total 163.000 163
a. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10)
b. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10), Gender
c. Predictors: (Constant), Zscore: Attractiveness (Mean Rating out of 10), Gender, Interaction Variable
d. Dependent Variable: Zscore: Support (from -1 to 1)

19
 11/10/2018

MODERATOR ANALYSIS Coefficientsa

Standardiz
ed
Unstandardized Coefficient Collinearity
Coefficients s Correlations Statistics
Zero- Toleran
Model B Std. Error Beta t Sig. order Partial Part ce VIF
1 (Constant) 6.563E- .078 .000 1.000
17
Zscore: -.029 .079 -.029 -.367 .714 -.029 -.029 -.029 1.000 1.000
Attractiveness
(Mean Rating out
of 10)
2 (Constant) -.057 .111 -.514 .608
Zscore: -.028 .079 -.028 -.361 .719 -.029 -.028 -.028 1.000 1.000
Attractiveness
(Mean Rating out
of 10)
Gender .114 .157 .057 .727 .468 .057 .057 .057 1.000 1.000
3 (Constant) -.055 .107 -.512 .609
Zscore: -.324 .109 -.324 -2.974 .003 -.029 -.229 -.225 .482 2.076
Attractiveness
(Mean Rating out
of 10)
Gender .114 .151 .057 .755 .451 .057 .060 .057 1.000 1.000
Interaction .570 .151 .411 3.769 .000 .177 .286 .285 .482 2.076
Variable
a. Dependent Variable: Zscore: Support (from -1 to 1)

MODERATOR ANALYSIS

Y = a + b1(X) + b2(M) + b3 (X*M)

a b1 b2 b3 X M Y
(Attractiveness) (Gender)
-0.055 -0.324 0.114 0.57 1 0 -0.379
-0.055 -0.324 0.114 0.57 0 0 -0.055
-0.055 -0.324 0.114 0.57 -1 0 0.269
-0.055 -0.324 0.114 0.57 1 1 0.305
-0.055 -0.324 0.114 0.57 0 1 0.059
-0.055 -0.324 0.114 0.57 -1 1 -0.187

20
 11/10/2018

MODERATOR ANALYSIS

MODERATOR ANALYSIS

These results tell us that the relationship

between attractiveness of a person and amount
of support given to their spouse is different for
men and women. Specifically, for women, as
attractiveness increases the level of support
that they give to their husbands increases,
whereas for men, as attractiveness increases
the amount of support they give to their wives
decreases

21
 11/10/2018

MODERATOR ANALYSIS
• Using Hayes’ PROCESS
Choose Model
Analyze Number (1 – for “OK”
basic moderation)

Plot data using

Choose Moderator
Regression GLM Univariate
Variable (M)
Function

Choose
Linear Independent
Variable (X)

Choose Outcome
PROCESS
Variable (Y)

MODERATOR ANALYSIS
Run MATRIX procedure:

************* PROCESS Procedure for SPSS Release 2.16.1 ******************

Written by Andrew F. Hayes, Ph.D. www.afhayes.com

Documentation available in Hayes (2013). www.guilford.com/p/hayes3

**************************************************************************
Model = 1
Y = Support
X = Attract
M = Gender

Sample size
164

**************************************************************************
Outcome: Support

Model Summary
R R-sq MSE F df1 df2 p
.292 .085 .041 4.975 3.000 160.000 .003

Model
coeff se t p LLCI ULCI
constant .233 .016 14.755 .000 .202 .265
Gender .024 .032 .755 .451 -.039 .086
Attract -.007 .014 -.513 .609 -.035 .020
int_1 .105 .028 3.769 .000 .050 .161

22
 11/10/2018

MODERATOR ANALYSIS
Product terms key:

int_1 Attract X Gender

R-square increase due to interaction(s):

R2-chng F df1 df2 p
int_1 .081 14.209 1.000 160.000 .000

*************************************************************************

Conditional effect of X on Y at values of the moderator(s):

Gender Effect se t p LLCI ULCI
-.500 -.060 .020 -2.974 .003 -.100 -.020
.500 .046 .019 2.346 .020 .007 .084

Values for quantitative moderators are the mean and plus/minus one SD from mean.
Values for dichotomous moderators are the two values of the moderator.

MODERATOR ANALYSIS
Data for visualizing conditional effect of X on Y
Paste text below into a SPSS syntax window and execute to produce plot.

DATA LIST FREE/Attract Gender Support.

BEGIN DATA.

-1.137 -.500 .290

.000 -.500 .222
1.137 -.500 .154
-1.137 .500 .194
.000 .500 .245
1.137 .500 .297

23
 11/10/2018

MODERATOR ANALYSIS
Output from PROCESS

MODERATOR ANALYSIS
• Sample Interpretation
Moderator Analysis was conducted to test whether the relationship
between Attractiveness of Spouse and Support for the other significantly
differs between husbands and wives. Moderation is shown by a
significant interaction effect, and in this case the interaction is highly
significant, b = 0.105, t=3.769, p < .01, indicating that the relationship
between attractiveness and support is moderated by gender.

Specifically, the male gender has shown a significant negative relationship

between attractiveness and support, b = −0.060, t=−2.95, p < .01. 2.
Whereas for females, there is a significant positive relationship between
attractiveness and support, b = 0.05, t = 2.12, p < .05. Furthermore, the
interaction between Attactiveness and Gender accounted for 8.1% of
variance in the criterion, Support (R2=.081, F(1,160)=14.209, p=<.001),
whereas the combination of the predictor, moderator and their
interaction explain 8.5% of the variance in in the criterion

24
 11/10/2018

MODERATOR ANALYSIS
• Sample Conclusion
Based on the analysis, we can conclude that
the relationship between attractiveness and
support is positive for wives (more attractive
wives give their husbands more support), but
negative for husbands (more attractive
husbands give their wives less support than
unattractive ones)

ASSIGNMENT 8
(data: McNulty et.al.(2008).sav)
1. Using PROCESS, run a mediation analysis with
Relationship Satisfaction as mediator between
Attractiveness (IV) and Support (DV).
a. Interpret results
b. Paste all necessary tables
2. Using PROCESS, test whether Gender is a significant
moderator between Attractiveness (IV) and Relationship
Satisfaction
a. Interpret results
b. Generate graph
c. Paste all necessary tables

Deadline (Nov. 16, 5pm)

25