How respondents view the advertisement of three brands is assumed to be influenced by their attitudes toward the brands and their demographics. Therefore, the brand index score, which represented respondents’ attitude toward each brand, was the first to be examined. Next, the relations between each brand’s brand index score and pre-to-post ad exposure constant sum score, which represented how the respondents view each advertisement, was examined as well.

1. Paired t-tests

Table 1. Brand Index Score

Number of Respondents = 73

Table 2. Paired t-test of Brand Index Score

*p < .15, Number of Respondents = 73

The result of the correlated t-test shows that the difference between mean brand index score for the test on Evian and Fiji is statistically significant at an alpha level of .15. In 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is expected that the mean brand index scores for Evian and Fiji would be about what they are in this sample. Therefore, we can project the results for this sample to the population; the same as for the tests on Evian and Aquafina. However, the difference between the mean brand index score for Fiji and Aquafina would not have a significantly different magnitude as found in this sample.
Overall, the respondents had the most positive perception for Fiji, the second for Aquafina with an almost similar mean score, and the third for Evian.

2. Between-Groups t-tests

Table 3. Brand Index Score Change for Evian

*p < .15, Total Cell Count: 52 Respondents

Table 4. Ad Index Score Change for Evian

*p < .15, Total Cell Count: 52 Respondents

There were 25 individuals (34.3% of the sample) who changed scores that moved up pre-to-post exposure to the advertisements on Evian. The mean of brand index score for this group was 32.2 and the mean of ad index score was 8.6. There were 27 individuals (37.0% of the sample) who changed scores that moved down pre-to-post exposure to the advertisements on Evian. The mean of brand index score for this group was 31.5 and the mean of ad index score was 4.9.

Respondents who moved up in pre-to-post exposure to the advertisement liked the Evian brand and advertisement more than respondents who moved down. Those who were more favorable toward the Evian brand after viewing the advertisement gave a significantly higher rating to the brand than those who were less favorable toward Evian after viewing the advertisement. Those who were more favorable toward the Evian brand after viewing the advertisement gave a much higher rating to the advertisement than did those who were less favorable toward Evian after viewing the advertisement.

At an alpha level of .15, the ad index score for Evian was statistically significant, but the brand index score for Evian was not. In 85 or more samples out of every 100 samples drawn from the same population as this sample, it is expected that the mean scores for Evian on its Ad index score would be about what they are in this sample. Thus, the results for this sample can be projected to the population. However, this does not hold true for the mean scores for Evian on its brand index score. The results for this sample can not be projected to the population.

3. Chi-Squared Significance Test

Table 5. Chi-Square Test for Evian

Chi-Squared=12.15*, Total Cell Count: 67 Respondents
*p < .15

The median for Evian’s brand index score is 30.0. From the sample of 73 respondents, 40.3% of the respondents felt less favorably about Evian after viewing the advertisement. Out of the 40.3% respondents, 63.0% rated the brand above the median and 37.0% rated the brand below the median.

With a chi-squared value of 12.15, it is expected to see the same distribution when 85 or more samples out of every 100 samples are drawn from the same population as this sample. Therefore, we can project the results for this sample to the population.

4. Frequency

Table 6. Pre-to-Post Exposure Measurement on the Constant-Sum Scale

While Fiji had the greatest increase in favorability (33 respondents or 45.2% of the sample) in pre-to-post exposure, Aquafina had the least increase in favorability (18 respondents or 24.7% of the sample) in pre-to-post exposure. Evian was the second highest moving up in favorability (25 respondents or 34.3% of the sample) in pre-to-post exposure. These results indicate that respondents favor Fiji’s advertisement the most, Evian’s advertisement the second, and Aquafina’s advertisement the least.

5. Frequency Counting

Out of the 73 respondents, 24 individuals (32.9% of the sample) rated Evian higher than Fiji, and 49 individuals (67.1% of the sample) rated Evian lower than or equal to Fiji after viewing the advertisement. The result shows that more respondents felt more favorably toward Fiji than toward Evian after viewing the advertisement.

6. Simple Correlation Coefficient

The simple correlation coefficient between brand index scores for Evian and Fiji is .2. Although there is a weak positive linear relationship between brand index scores for Fiji and Aquafina, this is statistically significant at an alpha level of .15. In 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is expected that the correlation between brand preferences for Fiji and Aquafina would be about the same as found in this sample. Therefore, we can project the results of the survey to the population. 

7. Simple Correlation Coefficient Modified

A simple correlation coefficient test was conducted between brand index score and ad index score for Aquafina to determine if any relationship exists among only the respondents who think taste is a key factor to decide to purchase bottled water. Twenty people out of the total sample of seventy three were used in this analysis.

To find out the relationship between each brand attribute and the pre-to-post ad exposure change score, multiple regression tests were conducted for each brand with the brand attributes as the independent variables and the pre-to-post ad exposure change score as the dependent variable.

<Multiple Regression Analysis for Evian>

Table 1. Correlation between the Brand Attribute and the Change Score

*p < .15

Table 2. Unstandardized and Standardized Regression Coefficients

*p < .15

With the R squared of 16.7%, the multiple regression analysis for Evian shows that there is a low relationship between how respondents rated the Evian brand on the ten Likert items regarding brand attributes and their favorability of the Evian advertisement as indicated by their pre-to-post ad exposure change score. Differences in the ten Likert items account for only 16.7% of the variation in the pre-to-post ad exposure change scores. Standard error of estimate is high at 2.1, indicating respondents were an average of 2.1 units away from the regression line. The F-ratio of 1.24 shows that in 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is not expected that the coefficient of multiple determinants would be about what it is in this sample. Therefore, we cannot project the results of the survey to the population.

The following offers the multiple regression equation:
Pre-to-Post ad Exposure Change Score for Evian = -2.3a +.3(A Good Brand) - .2(Don’t Like Taste) + .2(Too Expensive) - .3(Feel Healthy) + .5(A Trust Brand) + .4(For My Caloric Intake) + .6(Purer Water Source) - .2(Would Not Drink) + .2(Like Any Other Brand) - .9(Fits My Lifestyle). 

The brand attributes that are important independent variables in explaining variance in the change score in this sample include those described as “A Trust Brand”, “Purer Water Source”, and “Fits My Lifestyle”.

<Multiple Regression Analysis for Fiji>

Table 3. Correlation between the Brand Attribute and the Change Score

*p < .15

Table 4. Standardized and Unstandardized Regression Coefficients

*p < .15

There is a low relationship between the ten Likert items regarding brand attributes for Fiji and their pre-to-post ad exposure change score for Fiji. Differences in the ten Likert items account for only 16.7% of the variation in the pre-to-post ad exposure change scores. Standard error of estimate is high at 1.9, indicating respondents were an average of 1.9 units away from the regression line. The F-ratio of 1.24 shows that in 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is not expected that the coefficient of multiple determinants would be about what it is in this sample. Therefore, the results of the survey cannot be projected to the population.
The brand attributes that are important independent variables in explaining variance in the change score in this sample include those described as “A Good Brand”, “Feel Healthy”, “A Trust Brand”, “Would Not Drink”.

<Multiple Regression Analysis for Aquafina>

Table 5. Correlation between the Brand Attribute and the Change Score

*p < .15

Table 6. Standardized and Unstandardized Regression Coefficients

*p < .15

There is a low relationship between how respondents rated the Aquafina brand on the ten Likert items regarding brand attributes and their favorability of the Aquafina advertisement as indicated by their pre-to-post ad exposure change score. Differences in the ten Likert items account for only 15.0% of the variation in the pre-to-post ad exposure change scores. Standard error of estimate is high at 2.4, indicating respondents were an average of 2.4 units away from the regression line. The F-ratio of 1.09 shows that in 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is not expected that the coefficient of multiple determinants would be about what it is in this sample. Therefore, the results of the survey cannot be projected to the population.

In order to know if the ten Likert items can effectively predict our respondents’ attitude toward the advertisement with a better than chance accuracy, discriminant analysis for Evian was conducted. This discriminant analysis was executed to test whether people like the advertisement can be based on the ten Likert items of the brand with two groups, ‘up-movers’, those who liked the advertisement for Evian and ‘down-movers’, those who did not like the advertisement for Evian. Those respondents who indicated no change in their pre-to-post ad exposure brand evaluations were not included in this analysis. The independent variables in this analysis are the brand attributes as indicated by ten Likert items and the categorical dependent variable is group membership (up’s and down’s). Fifty two people out of the total sample of seventy three were used in this analysis.

Table 1. Group Mean Scores and Standard Deviation

When comparing the mean scores of each brand attribute, there are little differences between those who had positive pre-to-post ad exposure change scores and those who had negative pre-to-post ad exposure change scores. Although the largest difference between mean scores for up-movers and down-movers was 0.4, the difference was not great. The most marked appeared in the attributes “Purer Water Source”, “Would Not Drink”, and “Fits My Lifestyle”. Since there are no items that have dramatically different mean scores between up-movers and down-movers, there does not seem to be much to discriminate.

Table 2. Standardized and Unstandardized Discriminant Function Coefficients

The item “Fits My Lifestyle” has the biggest absolute value of standardized discriminant function coefficients, which is 1.2. We assume that the items with the absolute value of the standard discriminant function coefficients that equals or is larger than .6 (=1.2/2) are important independent variables in explaining differences in pre-to-post ad exposure change scores between up-movers and down-movers.

The important variables in discriminating between up-movers and down-movers are brand attributes described as “A Good Brand”, “A Trust Brand”, “For My Caloric Intake”, “Purer Water Source”, and “Fits My Lifestyle”.  The more people think Evian is a good brand, the more they like the advertisement; the more people think Evian is a brand they trust, the like the advertisement; the more people think they drink Evian because they watch their caloric intake, the more people like the advertisement; the more people think Evian’s water source is purer than that of other water, they like the advertisement. The more people think drinking Evian fits their lifestyle, however, the less they like the advertisement.

Table 3. Group Centroids and Statistical Significance

*p < .15, Degrees of Freedom=10

The group centroids, which are the average discriminant z scores of .7 for up-movers and of -.6 for down-movers, are significant, due to the statistical significance found for Wilks’ Lambda at .70 and chi-squared at 15.98. Therefore, in 85 or more samples out of every 100 samples drawn from the same population as this sample of 52 people, it is expected that the group centroids would be about what it is in this sample. Therefore, we can project the results of the survey to the population.
 Table 4. Classification Matrix

% of Classification Accuracy=69.2%

The following offers t-ration for observed data:

* p < .15, t₀ > t_c

The classification matrix shows that 69.2% (=36/52x100%) of the actual groups are predicted correctly. Out of 25 actual up-movers, 16 are correctly predicted to move up, and out of 27 actual down-movers, 20 are correctly predicted to move down.

If t₀ > t_c, then the result is significant. Since our observed t-ratio equals 2.03 and is more extreme than the critical t-ratio of 1.04, t-ratio to test classification accuracy is significant. Therefore, in 85 or more samples out of every 100 samples drawn from the same population as this sample of 52 people, it is expected that the percentage of classification accuracy would be about what it is in this sample. Therefore, the results of the survey can be projected to the population.

This is a quite successful discriminant analysis. (1) Even though there are little mean differences of each brand attribute between up-movers and down-movers, (2) Wilks’ Lambda result shows that group centroids are statistically significant as indicated by chi-squared value of 15.98. (3) Also, 69.2% of the original grouped cases were correctly classified as either an up-movers group or a down-movers group. Moreover, t-ratio to test this classification accuracy is significant. Although these ten brand attributes do not discriminate these two groups (up- and down-movers) at a high level, there is a weak connection between group membership (up’s and down’s) and the perception of brands indicated by ten Likert items. Thus, respondents’ attitudes toward the Evian brand can affect whether or not they like the advertisement for Evian.       

Using a brand attribute, “Drinking Aquafina fits my lifestyle” as a dependent variable, Analysis of variance (ANOVA) was executed to know the relationship with two categorical independent variables: (1) the household income level and (2) the likeability of the Aquafina print advertisement based on the change score direction. The respondents’ income level was split into two levels: “high” ($50,000 +) and “low” ($0 - $49,999). And then, multivariate analysis of variance (MANOVA) took all of the ten brand attributes into account as dependent variables.

<ANOVA for Aquafina>

Table 1. Mean Scores, by Household Income Level, for Respondents Who
Moved Up, Down, or Stayed the Same on Change Score for “Drinking Aquafina fits my lifestyle” Variable

Among respondents with both high and low income, the brand attribute score for “Drinking Aquafina fits my lifestyle” is highest for up-movers and lowest for non-movers. The difference between high income and low income of respondents is biggest for non-movers and smallest for up-movers. However, in general, the mean scores of the change score direction and respondents’ income level are relatively similar in value.

Table 2. Group Effects for Change Score and Household Income Level: ANOVA
Table

*p < .15

Of three effects, only the main effect for the change score was significant at an alpha level of .15. This means that in 85 or more samples out of every 100 drawn from the same population, the expected mean scores on this Likert item, “Drinking Aquafina fits my lifestyle”, would be the same as in this sample of 73 people. The other main effect for income level and the interaction of the two, however, were not significant. This means that these mean scores apply only to this sample and do not project to the general population. Therefore, it is hard to find the relationship between the likeability of the Aquafina print advertisement and the income level on the brand attribute of “Drinking Aquafina fits my lifestyle.”

<MANOVA for Aquafina>

Table 3. Mean Scores, by Household Income Level, for Respondents Who
Moved Up, Down, or Stayed the Same on Change Score for ten Likert Items

Table 4. Group Effects for Change Score and Household Income Level:
MANOVA Table

*p < .15

A factor analysis was performed for each bottled water brand to place related brand attributes into independent groups (factors) in order to calculate a brand attitude score. The variables used were all ten Likert items. A paired t-test was then executed to determine the statistical significance for the mean differences between brand attitude scores.

Table 1. Total Variance Explained

In the case of Evian, three of ten factors had Eigenvalues greater than 1.0, indicating that they explained the variance of at least a single Likert item. These three factors combined explained 69.5% of the variance in the original ten Likert items (30.5% of variance is unexplained). In the case of Fiji, three factors combined accounted for 67.0% of the variance in the original ten Likert items. For Aquafina brand, the results are more powerful. Four factors combined accounted for 73.1% of the variance in the original ten Likert items.

Table 2. Communalities (h²) and Factor Loadings for Evian

A squared factor loading is the percentage of variance explained by a factor in original variables such as ten Likert items. If we add up the squared factor loading over all the factors, the result becomes communality (h²). Communalities above, for instance, show that 80% of the variance in each of the following – “A Trust Brand”, “For My Caloric Intake” – is explained by all factors.

The Factor Matrix was used in this analysis because it had the least amount of ambiguous rows that had more than one factor greater than or equal to the absolute value of .5. A positive item to evaluate brand attitudes, “A Good Brand” loaded on factor I. Seven out of ten brand attributes also loaded on factor I and had an absolute value above .5.

Table 3. Communalities (h²) and Factor Loadings for Fiji

The Factor Matrix was used in this analysis because it had fewer problems than the Varimax Rotated Matrix. A positive item to evaluate brand attitudes, “A Good Brand” loaded on factor I. Eight out of ten brand attributes also loaded on factor I and had an absolute value above .5.

Table 4. Communalities (h²) and Factor Loadings for Aquafina

Ambiguity, as one item loaded equally into two factors, was founded in the Varimax Rotated Matrix. Thus, the Factor Matrix was chosen for this analysis. A positive item to evaluate brand attitudes, “A Good Brand” loaded on factor I. Six out of ten brand attributes also loaded on factor I and had an absolute value above .5.

Table 5. Attitude Scores

Number of Respondents = 73

The attitude score reveals that respondents evaluated Aquafina (attitude score of 3.5) the most favorably followed by Fiji (attitude score of 3.4) and Evian (attitude score of 3.3).

Table 6. Paired t-test for Attitude Scores

*p < .15

The paired t-test shows that in 85 or more samples out of every 100 samples drawn from the same population as this sample of 73 people, it is expected that the difference between brand attitude scores for Evian and Aquafina would be about what they are in this sample. However, the magnitude of differences between brand attitude scores for Evian and Fiji and for Fiji and Aquafina cannot be projected to the population.