QUANTITATIVE RESEARCH METHOD

PLS-SPSS REPORT

Submitted By:

Submitted to: Sir Tehseen Jawaid

 RESEARCH: Raving Fans or Silent Critics? Brand Service Recovery's Impact on

Positive eWOM

Model

Consumer power

Business Service

Recovery Attitude Positive eWOM

Strategies

Independent: Business Service Recovery

Dependent: Positive eWOM

Moderator: Consumer power

Mediator: Attitude

Business Service Recovery: Service recovery refers to the response of the service provider

when a service becomes unavailable or stops functioning.

Positive eWOM: Electronic word-of-mouth (eWOM) interactions between companies and

consumers
Consumer power: Powerful customers may choose justice and the brand's apology above

financial restitution when it comes to service recovery.

Attitude: The study of attitudes has garnered significant attention and effort from researchers in

the field of advertising and promotional materials due to two primary factors.

1-INSTRUMENTATION SOURCE
Table 1. Questionnaire design and instrumentation source
Variables Items Sources

Business Service Recovery 5 Mostafa (2017)

Positive eWOM 5 Mostafa (2017)

Consumer power 4 Mostafa (2017)

Attitude 5 Mostafa (2017)

2. DESCRIPTIVE STATISTICS
Descriptive statistics describes the characteristics of data.
Table 2. Descriptive statistics.
Demographics Items Frequency Percentage

Gender Male 37 61.7

Female 23 38.3

Age Below 20 19 31.7

21-30 25 41.7

31-40 3 5.0

Above 40 13 21.7

Education Bachelor 20 33.3

Master 29 48.3
 Phd 11 18.3

No. of responses (n=60)

Interpretation
The gender shows that male was 61.7% and female was 38.3%. Moreover, the age bracket shows

that Below 20 was 31.7%, 21-30 was 41.7%, 31-40 was 5.00% and Above 40 was 21.7%. Lastly,

in education section Bachelor was 33.3%, Master was 48.3% and PHD was 18.3%.

3- CONVERGENT VALIDITY

Convergent validity takes two measures that are supposed to be measuring the same construct

and shows that they are related. Reliability Criteria

S.No Name Criteria

1 Cronbach’s Alpha > 0.7

2 rho_A > 0.7

3 Composite Reliability (CR) > 0.7

Convergent Validity and Criteria

S.No Name Criteria

1 Average Variance Explained > 0.5

(AVE)

2 Most of the factor loading in > 0.7

their respective construct

Table 3. Testing for reliability and convergent validity.

Construct Item loadings CB Alpha CR AVE
 Attitude Toward the Brand ATT1 0.824 0.922 0.941 0.763

ATT2 0.902

ATT3 0.918

ATT4 0.875

ATT5 0.843

Business Service Recovery Strategies BSRS1 0.830 0.719 0.841 0.639

BSRS2 0.797

BSRS4 0.771

Consumer power CP1 0.869 0.786 0.873 0.698

CP2 0.733

CP3 0.895

Positive eWOM WOM1 0.722 0.734 0.847 0.650

WOM2 0.870

WOM3 0.820

Note: CR = Composite Reliability, AVE = Average Variance Extracted

Interpretation
Convergent validity measures the extent to which related constructs (your outcome variable and

your covariates) are measured. It is estimated through AVE (average variance extracted) that

must be greater than 0.5 and factor loadings of greater than 0.7. As in Table 3, the established

cut-off points for convergent validity as proposed by [51,52] are identified by AVE values above

the threshold of 0.5; the factor loadings above 0.6. In order to clean items that did not meet

these criteria, R2, R4 and R6 were removed. In Table 3, the CR for all variables is greater than
0.7 satisfying a one of the requirements that variables must be reliable and consistent in their

measurement and variables must accurately represent the underlying construct.

4- DISCRIMINANT VALIDITY
Discriminant validity shows that two measures that are not supposed to be related are in fact,

unrelated. Both types of validity are a requirement for excellent construct validity.

S.No Name Criteria

1 Fornell and Larcker √AVE should be highest in row and

column

2 HTMT All values < 0.85 or < 1

3 Difference in factor loading >0.2 or > 0.1

4.1. Fornell & Larcker

Table 4. Discriminant validity: Fornell– Larcker criterion.

Construct ATT BSRS CP WOM

ATT 0.873

BSRS 0.650 0.799

CP 0.624 0.505 0.835

WOM 0.718 0.540 0.634 0.806

Interpretation

We evaluate the discriminant validity (DV)to confirm that a latent component is different (or

distinct) from other components in the model. To examine this, we compute the degree to which

each given construct meets the Fornell-Larcker criterion whereby the square root of average
variance extracted (AVE) must be greater than the highest correlation between any two

constructs.

4.2- Heterotrait-Monotrait Ratio (HTMT)

Results Table 5. Heterotrait– Monotrait ratio (HTMT) results.

Construct ATT BSRS CP WOM CP x ATT

ATT

BSRS 0.783

CP 0.727 0.658

WOM 0.858 0.722 0.775

CP x ATT 0.240 0.187 0.080 0.346

Interpretation

Another alternative is to consider the Heterotrait-Monotrait ratio (HTMT), which is below 0.85

for discriminant validity. Verifying that the components do not overlap provides evidence that

the model has structural integrity and its theoretical framework has clarity, and validates this

study.

4.3. Cross Loadings

The cross-loading value on each construct and other constructs should be varied by 0.1.

Items ATT BSRS CP WOM CP x ATT

ATT1 0.824 0.519 0.596 0.587 -0.197

ATT2 0.902 0.647 0.568 0.669 -0.248

ATT3 0.918 0.626 0.500 0.583 -0.300

ATT4 0.875 0.575 0.611 0.686 -0.181

 ATT5 0.843 0.452 0.440 0.604 -0.081

BSRS1 0.544 0.830 0.431 0.557 -0.241

BSRS2 0.434 0.797 0.225 0.378 -0.057

BSRS4 0.562 0.771 0.515 0.351 -0.082

CP1 0.594 0.483 0.869 0.556 -0.093

CP2 0.447 0.421 0.733 0.364 -0.012

CP3 0.519 0.384 0.895 0.622 -0.073

WOM1 0.462 0.316 0.304 0.722 -0.389

WOM2 0.653 0.422 0.594 0.870 -0.147

WOM3 0.598 0.544 0.579 0.820 -0.183

CP x ATT -0.235 -0.164 -0.078 -0.270 1.000

Interpretation

we found when we ran our model and the all showed greater loads on their Constructs with

differences greater than 0.1 and 0.2 between others constructs. This indicates that model has

met the crossover load requirements and has strong discriminant validity.

5. PREDICTIVE POWER OF CONSTRUCT

Predictive validity refers to the ability of a test or other measurement to predict a future outcome.

Here, an outcome can be a behavior, performance, or even disease that occurs at some point in

the future. In statistics, R² (R-squared) measures the goodness of fit of a model. A value close to

1 indicates a good model, meaning:

• The model explains a large portion of the variance in the data. • The predictions are close to the

actual values.
• The model is a good fit for the data.

Table 7. Predictive power of construct.

Construct R-square R-square adjusted Results

ATT 0.423 0.413 Moderate model

WOM 0.589 0.566 Good model

Criteria

• 0.00 - 0.30: Bad model (model explains less than 30% of variance)

• 0.30 - 0.50: Moderate model (model explains 30-50% of variance)

• 0.50 - 0.70: Good model (model explains 50-70% of variance)

• 0.70 - 1.00: Excellent model (model explains more than 70% of variance)

6. HYPOTHESIS TESTING

Table 8. Hypothesis testing.

Hypotheses Original Sample mean Standard T statistics (| P values Decision

sample (M) deviation O/STDEV|)

(O) (STDEV)

ATT -> WOM 0.489 0.482 0.110 4.434 0.000 Significant

BSRS -> ATT 0.650 0.661 0.071 9.157 0.000 Significant

CP -> WOM 0.319 0.335 0.115 2.764 0.006 Significant

CP x ATT -> -0.136 -0.123 0.078 1.749 0.080 Insignifican

WOM t
Interpretation

The analysis reveals that the relationships between ATT -> WOM, BSRS -> ATT as well as CP -

> WOM are statistically significant (p-value is less than 0.05), indicating that the hypothesis is

accepted. Lastly the moderating relationship between CP x ATT -> WOM is not statistically

significant (p-value > 0.05), indicating that the null hypothesis is accepted and there is no

substantial difference.

Reason of CP x ATT -> WOM is not statistically significant

The negative and insignificant relationship between consumer power and brand positive eWOM

(eWOM) might be predicted as the more consumer savy, the less brand positive eWOM

(eWOM) reviews. For example, an increase in consumer bargaining power will not influence

eWOM quality, as long as consumers have no incentives to publicize positive experiences. This

indicates that consumer influence is not the key driver of positive eWOM, but rather a range of

factors such as experience with the brand or product satisfaction.

7.1. One Sample T-Test

Table 9. One Sample T-Test.

One-Sample Test

Test Value = 0

95% Confidence Interval of

the Difference
Mean

t df Sig. (2-tailed) Difference Lower Upper

EWOM 26.358 59 .000 12.93333 11.9515 13.9152

Criteria:

 P Value < 0.05 Significant Hypothesis Is Rejected.

 P Value > 0.05 Insignificant Hypothesis Is Accepted.

As shown in Table 9 below, one sample t-test on dependent variable EWOM is rejected which

means that significant value is less than 0.05. This is a two-tailed test of hypothesis testing,

where the null hypothesis is that the population mean is equal to 0 and the alternative hypothesis

is that the population mean is not equal to 0 Hence, if we find the p-value to be less than the

significance level, we will reject the null hypothesis. Your average electronic word of mouth is

significantly greater than the test value (2) from your sample.

7.2. T-Test (Paired)

Table 10. Paired Sample T-Test.

Paired Differences

Variable Mean Std. Std. Error t Sig. (2- Result

Deviation Mean tailed)

Pair 1 ATT - Insignificant/

.20667 3.58414 .447 .657
EWOM Hypothesis Accepted

Criteria:

 P Value < 0.05 Significant Hypothesis Is Rejected.

 P Value > 0.05 Insignificant Hypothesis Is Accepted.

Table 10 states that dependent variable SP with paired sample t-test is accepted, because the

value of significance out of 0.05. Since the p-value is 0. 657, greater than 0.05, we cannot reject

the null hypothesis. This means that electronic word of mouth differences before and after the

intervention are not statistically significant.

7.3. T-Test (Independent Samples)

Table 11. Independent Sample T-Test.

Independent Samples Test

S.NO. Variable F t Sig. (2-tailed) Result

1 EWOM .605 213 .832 Insignificant/Hypothesis

Accepted

Criteria:

 P Value < 0.05 Significant Hypothesis Is Rejected.

 P Value > 0.05 Insignificant Hypothesis Is Accepted.

Interpretation

According to independent sample t-test applied to dependent variable SP shown in table 11 it is

accepted because the value of significance is higher than 0.05. The p-value (Sig. 0.832 (2-tailed),

which is greater than 0.05. This implies that the mean electronic word of mouth disparities of

both subsets have no statistical difference. Table 4 reveals that gender does not have an essential

impact on the dependent variable (electronic word of mouth differences).

8. ONE WAY ANOVA

One-way analysis of variance (ANOVA) is a statistical method for testing for differences in the

means of three or more groups.

Table 12. One Way ANOVA.

S. No. Variable Variance Sig. Result Mean Sig. Value Results

Value

1 EWOM 0.1 Insignificant 0.428 Insignificant because

because p>0.05/ Hypothesis

p>0.05 accepted (all means are

equal)

Criteria:

 P Value < 0.05 Significant Hypothesis is Rejected.

 P Value > 0.05 Insignificant Hypothesis is Accepted.

Interpretation

There was no significance effect between the independent variable (age) and the dependent

variable (electronic word of mouth) as per the one-way ANOVA table 12 so it was accepted that

the means of all group is equal.

8.1. Tukey

Test If equal variance is assumed then Tuckey test is applied to the dependent variable.

Multiple Comparisons

Dependent Variable: SP

Tuckey HSD
(I) Age Mean Difference (I- Sig. Results

J)

Below 20 21-30 Hypothesis accepted/all

-1.03158 .810
means are equal

31-40 Hypothesis accepted/all

.83509 .985
means are equal

Above 40 Hypothesis accepted/all

1.02996 .876
means are equal

21-30 Below 20 Hypothesis accepted/all

1.03158 .810
means are equal

31-40 Hypothesis accepted/all

1.86667 .853
means are equal

Above 40 Hypothesis accepted/all

2.06154 .396
means are equal

31-40 Below 20 Hypothesis accepted/all

-.83509 .985
means are equal

21-30 Hypothesis accepted/all

-1.86667 .853
means are equal

Above 40 Hypothesis accepted/all

.19487 1.000
means are equal

Above 40 Below 20 Hypothesis accepted/all

-1.02996 .876
means are equal
 21-30 Hypothesis accepted/all
-2.06154 .396
means are equal

31-40 Hypothesis accepted/all

-.19487 1.000
means are equal

Interpretation

The above table shows that since all the p-values of age are greater than 0.05, this indicates that

hypothesis is accepted and all the means are equal.

EWOM

Tukey HSD a, b

Age N Subset for alpha = 0.05

Above 40 13 11.7385

31-40 3 11.9333

Below 20 19 12.7684

21-30 25 13.8000

Sig. .703

Interpretation

Here, all means are equal as all of the values lies in the one column.

8.2. Games-Howell Test

If equal variance is not assumed then Games-Howells test is applied.

 Multiple Comparisons

Dependent Variable: EWOM

Games-Howell

(I) Age Mean Std. Error Sig. Result

Difference (I-

Below 20 21-30 Means are

-1.03158 1.21573 .831
equal

31-40 Means are

.83509 2.25561 .980
equal

Above 40 Means are

1.02996 1.17318 .816
equal

21-30 Below 20 Means are

1.03158 1.21573 .831
equal

31-40 Means are

1.86667 2.24480 .838
equal

Above 40 Means are

2.06154 1.15225 .296
equal

31-40 Below 20 Means are

-.83509 2.25561 .980
equal

21-30 Means are

-1.86667 2.24480 .838
equal

Above 40 .19487 2.22204 1.000 Means are

 equal

Above 40 Below 20 Means are

-1.02996 1.17318 .816
equal

21-30 Means are

-2.06154 1.15225 .296
equal

31-40 Means are

-.19487 2.22204 1.000
equal

9.1 Regression

9.1.1 For Gender and Age

Coefficients

Unstandardized Standardized

Coefficients Coefficients

Model B Std. Error Beta t Sig. Results

1 (Constant) 13.976 1.695 8.246 .000

Gender Negative
-.092 1.027 -.012 -.090 .929
effect/insignificant

Age Negative
-.422 .455 -.123 -.928 .357
effect/insignificant
a. Dependent Variable: EWOM

Interpretation

The regression results show that for gender, the effect on electronic word of mouth is negative

but insignificant, with a coefficient of -.092 (p-value = 0.929). For age, the effect is positive but

also insignificant, with a coefficient of -0.422 (p-value = 0.357). Therefore, gender does not have

a significant impact on electronic word of mouth in this model.

9.2 EViews Regression

Dependent Variable: EDUCATION

Method: Least Squares
Date: 01/14/25 Time: 22:45
Sample: 1 60
Included observations: 60

Coefficien
Variable t Std. Error t-Statistic Prob.

C 2.412830 0.306669 7.867859 0.0000

AGE -0.012229 0.082321 -0.148548 0.8824
GENDER -0.387712 0.185905 -2.085543 0.0415

R-squared 0.073604 Mean dependent var 1.850000

Adjusted R-squared 0.041099 S.D. dependent var 0.708902
S.E. of regression 0.694182 Akaike info criterion 2.156541
Sum squared resid 27.46763 Schwarz criterion 2.261258
Log likelihood -61.69622 Hannan-Quinn criter. 2.197501
F-statistic 2.264397 Durbin-Watson stat 1.268460
Prob(F-statistic) 0.113162

Interpretation of Results:

1. Model Summary:

o Dependent Variable: EDUCATION (measured as a binary or categorical

outcome, e.g., level of education attained).

 o Method: Least Squares regression, indicating a linear model has been applied.

o Sample Size: 60 observations included in the analysis.

2. Coefficients:

o Constant (C): The intercept coefficient is 2.412830, meaning that when AGE and

GENDER are zero, the predicted value of EDUCATION is 2.41. This is likely a

baseline level of education.

o AGE: The coefficient for AGE is -0.012229, indicating a negative relationship

between AGE and EDUCATION, but the effect is very small and statistically

insignificant (p = 0.8824 > 0.05). This suggests that age does not have a

meaningful impact on education in this model.

o GENDER: The coefficient for GENDER is -0.387712, which implies that being

male (assuming GENDER is coded as 1 for male and 0 for female) is associated

with a reduction in EDUCATION by 0.39 units compared to being female. This

result is statistically significant (p = 0.0415 < 0.05), indicating that gender has a

significant impact on education.

3. Goodness-of-Fit:

o R-squared: 0.073604 indicates that only 7.36% of the variation in EDUCATION

is explained by AGE and GENDER. This is a low value, suggesting that the

model does not explain much of the variation in education levels.

 o Adjusted R-squared: 0.041099 is slightly lower, accounting for the number of

predictors in the model. This further confirms that the model has limited

explanatory power.

o S.E. of regression: 0.694182 represents the standard error of the regression,

indicating the average distance of the observed values from the predicted values.

4. Statistical Tests:

o F-statistic: 2.264397 with a Prob(F-statistic) = 0.113162, indicating that the

overall model is not statistically significant at the 5% level. This suggests that

the independent variables, AGE and GENDER, collectively do not significantly

explain the variance in EDUCATION.

o Durbin-Watson stat: 1.268460, which is below 2, indicates potential positive

autocorrelation in the residuals. This may suggest some issues with the

independence of observations or omitted variables.

5. Model Issues:

o The low R-squared and high p-value for the F-statistic suggest that the model may

be under-specified (i.e., important variables influencing education may have been

omitted).

o The insignificant effect of AGE and low explanatory power highlight the need to

include more relevant predictors (e.g., socio-economic status, parental education,

etc.).
 o The significant negative effect of GENDER suggests that gender disparities may

exist in educational attainment, but further investigation is needed to determine

the underlying reasons.

CONCLUSION

Excellent customer service recovery strategies like speeding up issues resolution and offering

personalized customer help throw into significance satisfaction levels among customers. When

customers are satisfied with a product or service, they are likely to relay their experience to

others, which means that the chances of having positive electronic word-of-mouth (eWOM) are

increased. As a result, effective service recovery leads to a positive online reputation and eWOM

(Donthu et al., 2021).

Consumer power of the brand is significantly positive when companies have effective service

recovery of the brand (such as timely responses & satisfactory solutions). As a result, consumers

feel more satisfied and are able to trust you, thus improving brand trust and reputation. Put

differently, effective rehabilitation tactics might transform bad experiences into good ones, thus

reinforcing the strength of the brand.

There are no connections between the customer powers and the brand service recovery

strategies. The correlation, in fact, is negative and not statistically significant. This indicates that

entertaining more customer opinion does not have any effect on the brand's recovery attempts or

the customers' view towards the brand. The challenges the brand faces are immense, but

consumer demand for the brand remains (Peinkofer et al., 2022). As such, these attempts to

recover impact consumers' overall perception of the organization very little.

However, the negative and insignificant relationship could be expected the higher consumer

savvy, the lower brand positive eWOM (eWOM) review. For instance, consumer bargaining

power will not impact eWOM quality, unless consumers have no incentive to publicly disclose

positive experiences (Grewal et al., 2022). This suggests that factors outside of consumer

influence are the drivers of positive eWOM such as experience with the brand or satisfaction

with the product.

RECOMMENDATIONS

Lastly, if Pakistan based organizations intend to initiate favorable eWOM, they must focus on

establishing successful service recovery strategies. As part of this effort, we will be training our

customer care representatives to listen carefully and understand client concerns, to resolve issues

quickly, and large, yet fair compensation where appropriate. When brands show real care for

what consumers think and are willing to take the proper action, they can turn unhappy followers

into loyal consumers and an online mouthpiece for your brand.

Companies need to engage with customers on social media in order to effectively monitor and

manage their online reputation. Moreover, brands in Pakistan, given its significant social media

usage, can have benefits from having a substantial presence on platforms, as well as taking

address to both positive and negative comments quickly. A potential preventative for negative

attitude and an enhancer of positive eWOM is to solicit reviews from satisfied consumers,

promote endorsement of their favorable experiences on web reviews.

Appendices

Appendix A
 Standar
Origin Sampl d
al e deviati T statistics P
sampl mean on (|O/STDE value
e (O) (M) (STDEV) V|) s
0.00
ATT -> WOM 0.489 0.482 0.110 4.434 0
0.00
BSRS -> ATT 0.650 0.661 0.071 9.157 0
0.00
CP -> WOM 0.319 0.335 0.115 2.764 6
- 0.08
CP x ATT -> WOM -0.136 0.123 0.078 1.749 0

ATT BSRS CP WOM CP x ATT

ATT1 0.824
ATT2 0.902
ATT3 0.918
ATT4 0.875
ATT5 0.843
BSRS1 0.830
BSRS2 0.797
BSRS4 0.771
CP1 0.869
CP2 0.733
CP3 0.895
WOM1 0.722
WOM2 0.870
WOM3 0.820
CP x ATT 1.000
 R-square R-square adjusted
ATT 0.423 0.413
WOM 0.589 0.56
Cronbach's alpha Composite reliability (rho_a) Composite reliability (rho_c) Average variance extracted (AVE)
ATT 0.92 0.927 0.941 0.763
BSRS 0.719 0.723 0.841 0.639
CP 0.786 0.834 0.873 0.698
WOM 0.734 0.765 0.847 0.650

ATT BSRS CP WOM CP x ATT

ATT
BSRS 0.783
CP 0.727 0.658
WOM 0.858 0.722 0.775
CP x ATT 0.240 0.187 0.080 0.346

AT BSRS CP WOM
AT 0.873
BSRS 0.650 0.79
CP 0.624 0.505 0.835
WOM 0.718 0.540 0.634 0.806
 ATT BSRS CP WOM CP x ATT
ATT1 0.824 0.519 0.596 0.587 -0.197
ATT2 0.902 0.647 0.568 0.669 -0.248
ATT3 0.918 0.626 0.500 0.583 -0.300
ATT4 0.875 0.575 0.611 0.686 -0.181
ATT5 0.843 0.452 0.440 0.604 -0.081
BSRS1 0.544 0.830 0.431 0.557 -0.241
BSRS2 0.434 0.797 0.225 0.378 -0.057
BSRS4 0.562 0.771 0.515 0.351 -0.082
CP1 0.594 0.483 0.869 0.556 -0.093
CP2 0.447 0.421 0.733 0.364 -0.012
CP3 0.519 0.384 0.895 0.622 -0.073
WOM1 0.462 0.316 0.304 0.722 -0.389
WOM2 0.653 0.422 0.594 0.870 -0.147
WOM3 0.598 0.544 0.579 0.820 -0.183
CP x ATT -0.235 -0.164 -0.078 -0.270 1.000

One-Sample Test

Test Value = 0

95% Confidence Interval of

the Difference
Mean

t df Sig. (2-tailed) Difference Lower Upper

EWOM 26.358 59 .000 12.93333 11.9515 13.9152

Paired Samples Statistics

 Std. Std. Error

Mean N Deviation Mean

Pair 1 ATT 13.1400 60 5.13113 .66243

EWOM 12.9333 60 3.80074 .49067

Paired Samples Test

Paired Differences

95% Confidence Interval of

the Difference
Std. Error

Mean Std. Deviation Mean Lower Upper t d

Pair 1 ATT -
.20667 3.58414 .46271 -.71921 1.13255 .447 5
EWOM

Independent Samples Test

Levene's Test for Equality of t-test for Equality of Means

Variances
 F Sig. t df Sig. (2-tailed)

EWOM Equal variances assumed .605 .440 .213 58 .832

Equal variances not

.218 50.722 .828
assumed

ANOVA

EWOM

Sum of

Squares df Mean Square F Sig.

Between Groups40.855 3 13.618 .940 .428

Within Groups 811.438 56 14.490

Total 852.293 59

Multiple Comparisons
Dependent Variable: EWOM

Mean 95% Confidence Interval

Difference (I-

(I) Age (J) Age J) Std. Error Sig. Lower Bound Upper Bound

Tukey HSD Below 20 21-30 -1.03158 1.15855 .810 -4.0993 2.0361

31-40 .83509 2.36487 .985 -5.4268 7.0970

Above 40 1.02996 1.37013 .876 -2.5980 4.6579

21-30 Below 20 1.03158 1.15855 .810 -2.0361 4.0993

31-40 1.86667 2.32585 .853 -4.2919 8.0253

Above 40 2.06154 1.30162 .396 -1.3850 5.5081

31-40 Below 20 -.83509 2.36487 .985 -7.0970 5.4268

21-30 -1.86667 2.32585 .853 -8.0253 4.2919

Above 40 .19487 2.43816 1.000 -6.2611 6.6508

Above 40 Below 20 -1.02996 1.37013 .876 -4.6579 2.5980

21-30 -2.06154 1.30162 .396 -5.5081 1.3850

31-40 -.19487 2.43816 1.000 -6.6508 6.2611

Games-Howell Below 20 21-30 -1.03158 1.21573 .831 -4.2878 2.2246

31-40 .83509 2.25561 .980 -10.7373 12.4075

 Above 40 1.02996 1.17318 .816 -2.1616 4.2215

21-30 Below 20 1.03158 1.21573 .831 -2.2246 4.2878

31-40 1.86667 2.24480 .838 -9.8126 13.5459

Above 40 2.06154 1.15225 .296 -1.0527 5.1758

31-40 Below 20 -.83509 2.25561 .980 -12.4075 10.7373

21-30 -1.86667 2.24480 .838 -13.5459 9.8126

Above 40 .19487 2.22204 1.000 -11.7611 12.1509

Above 40 Below 20 -1.02996 1.17318 .816 -4.2215 2.1616

21-30 -2.06154 1.15225 .296 -5.1758 1.0527

31-40 -.19487 2.22204 1.000 -12.1509 11.7611

EWOM

Subset for

alpha = 0.05

Age N 1

Tukey HSDa,b Above 40 13 11.7385

31-40 3 11.9333

Below 20 19 12.7684

21-30 25 13.8000

Sig. .703

Means for groups in homogeneous subsets are

displayed.

a. Uses Harmonic Mean Sample Size = 7.954.

b. The group sizes are unequal. The harmonic

mean of the group sizes is used. Type I error

levels are not guaranteed.