Basic Statistics
Multiple Regression Analysis
Discriminant Analysis
ANOVA/MANOVA
Factor Analysis
Basic Statistics
Paired T-Test
|
Mean Brand Index Score |
Old Spice |
32.2 |
Lever 2000 |
32.3 |
Dial |
32.1 |
|
t-ratio |
Standard Deviation |
Sample Size |
Old Spice – Lever |
.25 |
5.92 |
66 |
Old Spice – Dial |
.04 |
5.91 |
66 |
Lever 2000 – Dial |
.53 |
3.27 |
66 |
The Brand Index score is a composite total of points based on the respondents answers to the ten Likert items in the survey. Each question had five response options ranging in value from 5 for the most favorable response about the brand to 1 for the least favorable response about the brand. Accordingly, Brand Index scores range from 10-50 and give insight into how favorably the respondents viewed the three brands. The paired t-test finds the difference between paired Brand Index scores and then calculates the mean. The t-ratio relates to the closeness of the means.
Lever has the highest brand index score at 32.3 followed closely by Old Spice at 32.2 and Dial at 32.1 respectively. All three are close in their brand index mean, which indicates parity amongst the three brands within the sample size. The low t-ratio is demonstrates how close all three brands were in their assessment from the respondents. The t-ratio for the three paired t-test were not statistically significant, which means that the data from this t-test cannot be confidently applied to the outside world.
Group t-tests
|
Brand Index Score Mean |
Ad Index Score Mean |
Sample Size |
Up movers with Old Spice |
32.4 |
7.0 |
16 |
Down movers With Old Spice |
32.2 |
5.8 |
17 |
Old Spice |
T-Ratio |
Brand Index Score |
.17 |
Ad Index Score |
1.34* |
*p≤.15
|
Brand Index Score Mean |
Ad Index Score Mean |
Sample Size |
Up movers with Lever 2000 |
33.3 |
6.9 |
26 |
Down movers with Lever 2000 |
28.9 |
3.1 |
9 |
Lever 2000 |
T-Ratio |
Brand Index Score |
.00* |
Ad Index Score |
.00* |
*p≤.15
The two group t-test was done between respondents who moved up and moved down in their point assessment for Old Spice and Lever 2000 after exposure to the advertisements. The dependent variables that we measured in the t-test were the Brand Index score and the Ad Index score for Old Spice and Lever 2000, while the dependant variable was whether the respondent moved up or down. The Ad Index is a composite total of points from respondents based on how they answered positively worded items related to the advertisements. The Ad index score range for both Old Spice and Lever 2000 is 0 to 14.
The Brand Index and Ad Index score for those who moved up were higher for both Old Spice and Lever 2000 relative to those who moved down. This phenomenon would make logical sense, since respondents who thought well of the advertisement are likely to think more highly of the brand throughout the survey and select items accordingly. Only in the case of the Brand Index score for Old Spice was the result between up mover and down movers close. The Brand Index for Old Spice t-test was the only one t-test that was not statistically significant. The Brand Index score for Lever 2000 and the Ad Index score for both Old Spice and Lever 2000 were statistically significant. This means that in that in 85 or more samples drawn from the same population, the mean score for the Lever 2000 Brand index and Lever 2000 and Old Spice Ad Index would be about the same as they were in the sample.
Chi-squared test
Median Brand Index Score for Old Spice was 32 |
|
Above Median |
Below Median |
Up |
Total: 7
Row Percentage: 50%
Column Percentage: 23.3%
Total Percentage: 12.3% |
Total: 7
Row Percentage: 50%
Column Percentage: 25.9%
Total Percentage: 12.3% |
Same |
Total: 15
Row Percentage: 51.7%
Column Percentage: 50%
Total Percentage: 26.3% |
Total: 14
Row Percentage: 48.3%
Column Percentage: 51.9%
Total Percentage: 24.6% |
Down |
Total: 8
Row Percentage: 57.1%
Column Percentage: 26.7%
Total Percentage: 14.0% |
Total: 6
Row Percentage: 42.9%
Column Percentage: 22.2%
Total Percentage: 10.5% |
Chi Squared Value |
Degrees of Freedom |
Sample Size |
.16 |
2 |
57 |
The Chi-Squared test assess the significance of the relationship between respondents moving from their pre to post advertisement brand assessment relative to the median score on the constant sum scale. The results were quite surprising in that respondents that moved down pre to post assessment had the highest tendency to be over the median. This is a bit counter intuitive, as one would expect those who lowered their brand assessment to be the one’s likely to have a higher percentage falling below the median. Conversely, for those respondents that moved up in their brand assessment pre to post test there was parity amongst those above and below the median, which once again was a counter intuitive result. One most likely would have conjectured that individuals who raised their brand index score would have an increased percentage of Brand Index scores above the median. There are many possible explanations for these surprising results and more studies would certainly help in determining the influential variable at work in this case.
Frequency Table
n=66 |
Old Spice |
Lever 2000 |
Dial |
Brand Index Score went up from Pre to Post Exposure |
16 |
26 |
11 |
Brand Index Score went down from Pre to Post Exposure |
17 |
9 |
26 |
Brand Index Score stayed the same Pre to Post Exposure |
33 |
31 |
29 |
The chart indicates that each of the three brands advertisements had a different effect on the respondents. In the case of Old Spice, exposure to its advertisement had a neutral effect in that about half the sample size stayed the same, almost a quarter of the sample size moved up, and almost a quarter of the sample size moved down. For Lever 2000, exposure to the advertisement had a positive effect on the respondents’ pre-to-post advertisement as 39.4% moved up and only 13.6% moved down. For Dial, exposure to the advertisement had a negative effect on the respondents’ pre-to-post advertisement as 39.4% moved down while only 16.7% moved up.
Brand Index score Comparison
The number of respondents that had a higher brand index score for Old Spice relative to Lever 2000 |
34 |
Percentage of the Sample Size that had a higher brand index score for Old Spice Relative to Lever 2000 |
51.5% |
Sample size |
66 |
In our study that included 66 individuals, we had 34 respondents that had a higher brand index score for Old Spice relative to Lever 2000. The brand index score was based on an individual’s response to 10 Likert items that gauged brand preference.
From out data we can tell that our sample population had a noticeable preference for Old Spice. The fact that 51.5% preferred Old Spice as a brand is significant, since 6.1% (4 respondents) of the sample population were neutral in their brand preference. This means that only 42.4% preferred Lever 2000 as a brand. The preference for Old Spice as a brand amongst our sample population is important to note in that it may have an influencing effect on respondents other selections in the survey.
Correlation Coefficient
Correlation Coefficient between the Brand Index score for Old Spice and Brand Index score for Lever 2000 |
-.21* |
Sample Size |
66 |
*p≤ .15
There is a weak negative correlation between the brand index score for Old Spice and the brand index score for Lever 2000 amongst respondents. The weak negative correlation means that there is a weak pattern in our study of respondents having an inverse preference for the two brands Old Spice and Lever 2000. The correlation between the two brands was statistically significant. This means that in 85 or more samples drawn from the same population as this sample of 66 people, the mean score for the correlation between Old Spice and Lever 2000 would be about the same as they are in this sample.
This result indicates that there is a slight tendency for people to favor either Old Spice or Lever 2000 inversely. A potential explanation would be brand loyalty. Individuals that like either Old Spice or Lever 2000 as a brand could tend to dislike the other brand as a result of their loyalty. This is just one of many plausible explanations for why there is an inverse relationship in preference for the two brands. The negative correlation is important to be aware of when assessing what factors influence brand preference relative to Old Spice and Lever 2000.
A Segmented Correlation Coefficient
Correlation Coefficient between the Brand Index score for Old Spice and Brand Index score for Lever 2000 when you remove all respondents who choose to not give the Laundry Detergent Bold any points on the purchasing scale. |
-.11 |
Sample Size |
44 |
When one calculates a correlation coefficient between the Brand Index score for Old Spice and the Brand Index score for Lever 2000 for our data set after removing all respondents who choose not to give the laundry detergent Bold any points in the on the purchasing scale in the questionnaire, one will find that there a very weak negative correlation. This means that there is a weak indication that people in this study who like Old Spice as a brand will not also like Lever 2000 as a brand. However, this finding was not statistically significant.
What we can draw from this study is that the individuals in this study who do not like the Laundry Detergent Bold tend to have negatively correlated preference relative to Old Spice and Lever 2000. This evident by the fact that the negative correlation coefficient, which was calculated above in Question 6, lowered when you remove all the respondents that choose not to give Bold any purchasing points. This figure is valuable to note when one is assessing how laundry brand preference affected individuals’ preference for body wash.
Back to the top
Multiple Regression Analysis
I ran a linear regression analysis on each of the three brands of body wash in my study to determine the relationship between respondents’ brand score movement pre to post advertisement exposure and their brand characteristic preferences based on their responses to ten likert items. From this analysis one will gain insight into which brand characteristics for each body wash are correlated with how an individual is likely to react to an advertisement for each of the three brands. One point to note in this analysis is that for the likert items with negative prompts, the respondents’ answers were scored on an inverse point scale. The most negative response to a negative prompt received the highest points. So for each item we are assessing a positive interpretation of the brand relative to a respondents’ brand score movement pre to post advertisement exposure.
Old Spice
r: |
r-squared: |
Standard Error of Estimate |
f-ratio |
0.4 |
18.8% |
1.7 |
1.27 |
Brand Characteristics:
Likert Items |
b:
Unstandardized Coefficient |
β:
Standardized
Coefficient |
t-ratio |
Constant |
-2.0 |
|
.86 |
Old Spice is a good brand of body wash |
-0.2 |
-0.1 |
.69 |
I like the smell of Old Spice |
-0.2 |
-0.1 |
.77 |
Old Spice is too expensive |
0.2 |
0.1 |
.65 |
Old Spice does a better job of getting me clean |
.06 |
0.0 |
.87 |
Old Spice helps the health of my skin |
-0.3 |
-0.1 |
.92 |
Old Spice is not better than other brands |
0.4 |
0.2 |
.25 |
Using Old Spice fits my lifestyle |
0.2 |
0.1 |
.46 |
The opposite sex likes the smell of Old Spice |
0.1 |
0.1 |
.28 |
It is not important to me to use Old Spice |
0.4 |
0.2 |
1.55* |
I prefer another body wash over Old Spice |
0.3 |
0.2 |
.95 |
*p≤.15
For Old Spice the coefficient of multiple determination, r-squared, tell us that only 18.8% of the variance of the change on the constant sum scale for Old Spice is explained by the ten likert items. This indicates that there is only a weak relationship between how respondents rated Old Spice as a brand through the ten likert items and how they responded to the advertisements, which was assessed by their change score pre to post advertisement exposure. The likert item that had the strongest relationship with a respondent’s change score was the prompt that “it is not important to me to use Old Spice.” This likert item had a high t-ratio of 1.55 which was statistically significant. This means that in 85 or more samples out of every 100 samples, drawn from the same population of 66 people, we would expect the t-ratio for the likert item to be the same as it is in this sample.
The regression equation written out for Old Spice: Constant for Old Spice= -2.0 - .01 (good) - .01 (smell) + 0.1 (expensive) + 0.0 (clean) – 0.1 (health) + 0.1(not better) + 0.1 (lifestyle) + 0.1(opposite sex) + 0.1 (not important) + 0.1(prefer)
The qualitative interpretation of this regression equation goes as follows: (1) The more respondents thought that Old Spice was a good brand of Body wash; the less they liked the advertisement. (2) The more respondents liked the smell of Old Spice; the more they liked the advertisement. (3) The more respondents thought that Old Spice was not too expensive, the more they liked the advertisement. (4) The more respondents thought that Old Spice got them clean; the more they liked the advertisement. (5) The more respondents thought that Old Spice helped the health of their skin; the less they liked the advertisement. (6) The more respondents thought Old Spice was a better brand; the more they liked the advertisement. (7) The more respondents felt that Old Spice fit with their lifestyle; the more they liked the advertisement. (8) The more respondents thought that the opposite sex likes the smell of Old Spice; the more they liked the advertisement. (9) The more respondents thought using Old Spice was important to them; the more they liked the advertisement. (10)The more respondents preferred Old Spice as a brand; the more they liked the advertisement.
Lever 2000
r: |
r-squared: |
Standard Error of Estimate |
f-ratio |
0.5 |
24.4% |
2.3 |
1.77* |
*p≤.15
Brand Characteristics:
Likert Items |
b:
Unstandardized Coefficient |
β
Standardized
Coefficient |
t-ratio |
Constant |
-2.5 |
|
.83 |
Lever is a good brand of body wash |
0.4 |
0.1 |
.63 |
I like the smell of Lever |
-0.1 |
0.0 |
.18 |
Lever is too expensive |
-0.3 |
-0.1 |
.62 |
Lever does a better job of getting me clean |
0.0 |
0.0 |
.87 |
Lever helps the health of my skin |
0.7 |
0.2 |
1.27 |
Lever is not better than other brands |
0.3 |
0.1 |
.52 |
Using Lever fits my lifestyle |
1.4 |
0.4 |
2.34* |
The opposite sex likes the smell of Lever |
-0.7 |
-0.2 |
1.15 |
It is not important to me to use Lever |
0.2 |
0.1 |
0.47 |
I prefer another body wash over Lever |
-0.5 |
-0.2 |
1.29 |
*p≤.15
Lever 2000 has a stronger r-squared value than Old Spice, meaning that there was a stronger relationship between respondents’ change score and their response to the likert items. However, an r-squared value of 25% still indicates only a weak relationship between the dependant and independent variable.
Many of the likert items were inter-correlated as only four of the variables (the ones bolded for their beta coefficient scores) had a strong influence on the regression between the two variables. In the case of Lever 2000 it seems that the best indicator of whether a respondent will likely be a mover in their brand index score is whether the brand fits with their lifestyle. The likert item is highly correlated with a statistically significant t-ratio of 2.34. So this relationship can be applied to the outside world of this survey. In 85 or more samples out of every 100 samples, drawn from the same population of 66 people, we would expect the t-ratio for the likert item to be the same as it is in this sample.
Dial
R: |
r-squared: |
Standard Error of Estimate |
f-ratio |
0.4 |
12.8% |
2.6 |
.81 |
Brand Characteristics:
Likert Items |
b:
Unstandardized Coefficient |
β:
Standardized
Coefficient |
t-ratio |
Constant |
0.4 |
|
.11 |
Dial is a good brand of body wash |
0.0 |
0.0 |
.05 |
I like the smell of Dial |
-0.3 |
-0.1 |
.51 |
Dial is too expensive |
-0.4 |
-0.1 |
.77 |
Dial does a better job of getting me clean |
-0.5 |
-0.1 |
.76 |
Using Dial helps the health of my skin |
1.3 |
0.3 |
1.83* |
Dial is not better than other brands |
0.1 |
0.0 |
.10 |
Using Dial fits my lifestyle |
-0.2 |
-0.1 |
.27 |
The opposite sex likes the smell of Dial |
0.6 |
0.2 |
1.14 |
It is not important to me to use Dial |
-0.2 |
-0.1 |
0.60 |
I prefer another body wash over Dial |
-0.5 |
-0.2 |
1.47* |
*p≤.15
Of the three brands, Dial has the weakest relationship between respondent’s change score and their response to the likert items. Only 12.8% of the variance of the change on the constant sum scale for Dial is explained by the ten likert items. This weak relationship is reflected in the f-ratio being only .81. However, this f-ratio is not statistically significant and therefore in 85 or more samples out of every 100 samples, drawn from the same population of 66 people, we would not expect the f-ratio to be the same as it is in this sample.
In Dial’s case there are three likert items (the three bolded items) that have a significant relationship with the respondent’s moving in their brand score pre to post advertisement exposure. The relationship between the likert items related to preference which has a t-ratio of 1.83 and which has a t-ratio of 1.47 are both statistically significant. Therefore, one will likely be able to see that same t-ratio outside of this survey. Based on their β-scores one can see that these two likert items and the likert item concerning the opposite sex had the most bearing on regression between the two variables.
Back to the top
Discriminant Analysis
In this report I conducted a discriminant analysis to assess the feasibility of being able to discriminate between respondents that are up and down movers in their brand score pre to post advertisement exposure on the basis of their response to ten likert items related to brand traits. I conducted the study using the results I received from respondents regarding the brand Old Spice and had a total population size of 33.
Old Spice
Brand traits (Likert Items) |
Up- Movers
n=16 |
Down- Movers
n=17 |
Standardized Discriminant Coefficient |
Unstandardized Discriminant Coefficient |
Mean |
Std. Dev. |
Mean |
Std. Dev. |
|
|
Old Spice is a good brand of body wash |
3.6 |
1.0 |
4.0 |
0.9 |
-0.0 |
-0.0 |
I like the smell of Old Spice |
3.5 |
1.0 |
3.9 |
1.3 |
-0.4 |
-0.3 |
Old Spice is too expensive |
3.1 |
0.5 |
2.8 |
1.1 |
0.1 |
0.1 |
Old Spice does a better job of getting me clean |
2.9 |
0.8 |
2.6 |
0.6 |
0.4 |
0.5 |
Old Spice helps the health of my skin |
2.8 |
0.6 |
3.1 |
0.7 |
-0.7 |
-1.1 |
Old Spice is not better than other brands |
2.9 |
0.8 |
2.9 |
0.9 |
-0.3 |
-0.3 |
Using Old Spice fits my lifestyle |
2.9 |
0.9 |
3.1 |
1.1 |
0.8 |
0.7 |
The opposite sex likes the smell of Old Spice |
3.4 |
0.8 |
3.6 |
1.0 |
0.5 |
0.6 |
It is not important to me to use Old Spice |
4.1 |
0.9 |
3.4 |
1.0 |
0.8 |
0.9 |
I prefer another body wash over Old Spice |
3.9 |
1.2 |
2.9 |
1.1 |
0.6 |
0.5 |
In comparing the mean scores between up-movers and down-movers for the Old Spice likert items only two items stand out as having a significant difference. For the prompt “It is important to me to use Old Spice” the up-movers had a substantially higher mean score (0.7 higher) relative to the down-movers. For the prompt “I prefer another body wash over Old Spice” there was the highest mean score difference (1.0 higher for the up-movers) between the up-movers and the down movers. Both of these likert items with significantly different mean scores were negatively worded prompts. One must remember that negatively worded likert items were scored inversely; therefore, in this situation up-movers as a whole disagreed with the negatively worded prompt more than down movers. A possible explanation for this phenomenon is that up-movers, who had recently raised their impression of the brand, were reluctant to agree with a negative prompt, since that would be inconsistent with their new impression.
The standard discriminant coefficient tells us how closely a respondent’s answer to one of the likert items relates to their movement pattern pre to post advertisement exposure. In this case there were seven brand characteristics (the bolded scores indicate importance) that were important in discriminating between up and down movers. These seven important likert items tell us that:
- The more that a respondent likes the smell of Old Spice, the more likely they are to be a down-mover
- The more that a respondent believes that Old Spice does a better job of getting them clean, the more likely they are to be an up mover.
- The more that a respondent believes that Old Spice help the health of their skin, the more likely they are to be a down mover.
- The more that a respondent believes that Old Spice fits their lifestyle, the more likely they are to be an up mover.
- The more that a respondent believes that the opposite sex likes the smell of Old Spice; the more likely they are to an up mover.
- The more a respondent believes that using Old Spice is important to them, the more likely they are to be an up mover.
- The more a respondent prefers Old Spice over other body washes, the more likely they are to be an up mover.
The interesting thing about this result is that the likert item with the highest coefficients was the one concerned with lifestyle. The Old Spice advertisement that respondents were exposed to was a lifestyle centered advertisement. Therefore, one can conclude that the advertisement succeeded in appealing to the demographic of body wash purchasers who are concerned with lifestyle issue when purchasing body wash.
Brand Score Movement |
Group Centroids |
Up-Movers |
0.8 |
Down-Movers |
-0.7 |
Wilkes Lamba |
Chi-Squared |
df |
0.61 |
12.87 |
10 |
The Group Centroid scores are reasonably far apart and therefore distinguishable from each other. This means that one will have a moderate degree of success predicting whether a respondent will be an up-mover or a down mover on the basis of their z-score. The z-score is calculated through the discriminant function with the equation z = a + b1x1 +…+ bnxn.
The wilkes lamba test is used to evaluate the significance of the group centroids. This test allows us to assess whether we would see the same centriod scores in the general public. Unfortunately the wilkes lamba score of .61 based which is based on the chi-square score of 12.87 was not significant at .15 level. This means that in 85 or more out of 100 samples, drawn from the same population as this one, we would not expect to find group centroid values of the same magnitude.
|
Predicted |
|
t-ratio |
Actual |
Up-Movers |
Down-Movers |
Total |
1.89* |
Up-Movers |
12 |
4 |
16 |
Down-Movers |
4 |
13 |
17 |
|
|
|
33 |
|
75.8% of original grouped cases correctly classified. *p≤.15 |
T-Ratio= (.758-.242)/√((.758)(.242)/33)-((.5)(.5)/(33)= 1.89
The classification matrix above shows that based on brand preference as assessed by a respondents answers to the likert items that 12 out of 16 up mover were correctly assigned and 13 out of 17 down movers were correctly assigned. This gave us a correct discrimination total of 75.8%, which is quite an accurate figure given the small sample size that we were working with in this study. What this results means is that if one was given solely a respondent’s answers to the ten likert items, then one would have a 75.8% success rate of correctly discriminating whether they were an up-mover or down-mover in their brand score pre to post advertisement exposure. This result had a t-ratio of 1.89, which means that it is statistically significant. Therefore, in 85 or more out of a 100 samples, drawn from the same population as this one, we could expect to find the same success rate of discrimination as we find in this study.
Back to the top
ANVOVA
In my first analysis I conducted a two-way factorial ANOVA test to determine the relationship between the likert item which states “Old Spice is a good brand of body wash” and the movement pattern of respondents pre to post advertisement and the respondent’s gender. In this ANOVA analysis the likert item was the dependant variable while the respondent’s move score and gender were the two independent variables.
|
Up |
Same |
Down |
Male |
Mean = 3.8
Std. Dev. = 1.1
N= 10 |
Mean = 4.0
Std. Dev. = 0.9
N= 22 |
Mean = 4.2
Std. Dev. = 0.8
N= 14 |
Female |
Mean = 3.3
Std. Dev. = 0.5
N= 6 |
Mean = 3.0
Std. Dev. = 0.4
N= 11 |
Mean = 3.0
Std. Dev. = 0.0
N= 3 |
|
Sum of Squares |
Degrees of Freedom |
Mean Square |
F-Ratio |
Gender |
8.60 |
1 |
8.60 |
3.83* |
Change Score for Old Spice |
0.16 |
2 |
0.08 |
0.12 |
Change Score for Old Spice by Gender |
0.96 |
2 |
0.48 |
0.71 |
Error |
40.24 |
60 |
|
|
*P ≤ .15
From the ANOVA analysis one can see that for men there is a pattern of respondents’ having a higher mean score for the likert item as their movement score goes downwards. The difference between cross sections is evenly distributed, as each section has a .2 higher mean score. However, one should be aware that the .4 difference in mean scores between up-moving and down-moving males is still within the standard deviation for both cross sections. Therefore, one could expect to see results for each cross section that defy this pattern.
For females the up-movers generally had a higher mean score for the likert item. This result, unlike the male result, is what one would expect before seeing the results. It would seem logical that respondents who liked the advertisement, which would be reflected by a positive change score, would have a higher mean score for how highly they regarded the brand.
In this case only the relationship between the gender of a subject and their mean score was statistically significant. In 85 or more samples out of every 100 samples, drawn from the same population of 66 people, we would expect the same F-Ratio between the likert item and the gender of the respondent. This relationship between gender and their mean score could be applied to the general population. Unfortunately, the F-Ratio between the change score and the likert item and the F-Ratio between both independent variables and the likert item were not statistically significant. This means that the pattern we observed for both men and women and their change score relative to the likert item can not be applied to the general population.
MANOVA
Old Spice
Brand Traits
(Likert Items) |
Male |
Female |
Up
N=10 |
Same
N=22 |
Down
N=14 |
Up
N=06 |
Same
N=11 |
Down
N=03 |
mean |
s.d. |
mean |
s.d. |
mean |
s.d. |
mean |
s.d |
mean |
s.d. |
mean |
s.d |
Old Spice is a good brand of body wash |
3.8 |
1.1 |
4.0 |
0.9 |
4.2 |
0.8 |
3.3 |
0.5 |
3.0 |
0.4 |
3.0 |
0.0 |
I like the smell of Old Spice |
3.2 |
0.4 |
3.2 |
0.5 |
2.9 |
0.3 |
3.8 |
1.0 |
3.6 |
0.9 |
3.3 |
1.2 |
Old Spice is too expensive |
3.0 |
0.5 |
3.2 |
0.5 |
2.8 |
1.2 |
3.3 |
0.6 |
3.2 |
0.6 |
3.0 |
0.0 |
Old Spice does a better job of getting me clean |
3.0 |
0.9 |
3.2 |
0.5 |
2.6 |
0.6 |
2.7 |
0.5 |
2.4 |
0.7 |
2.3 |
0.6 |
Old Spice helps the health of my skin |
3.0 |
0.0 |
2.6 |
0.7 |
3.2 |
0.4 |
2.8 |
0.6 |
2.9 |
0.8 |
3.2 |
0.7 |
Old Spice is not better than other brands |
3.0 |
0.9 |
2.9 |
0.9 |
3.0 |
1.0 |
2.7 |
0.5 |
2.4 |
0.7 |
2.3 |
0.6 |
Using Old Spice fits my lifestyle |
3.1 |
0.3 |
2.9 |
0.3 |
3.0 |
0.0 |
2.3 |
0.5 |
1.9 |
0.9 |
1.3 |
0.6 |
The opposite sex likes the smell of Old Spice |
3.2 |
0.9 |
3.3 |
1.1 |
3.8 |
0.9 |
3.7 |
0.5 |
3.2 |
1.0 |
2.7 |
1.2 |
It is not important to me to use Old Spice |
4.1 |
0.6 |
3.7 |
1.0 |
3.2 |
0.8 |
4.0 |
1.3 |
3.7 |
1.3 |
4.0 |
1.7 |
I prefer another body wash over Old Spice |
3.4 |
1.2 |
2.8 |
0.9 |
2.6 |
0.8 |
4.7 |
0.8 |
4.4 |
0.7 |
4.7 |
0.6 |
In this second analysis I did a two-way factorial MANOVA test to determine the relationship between the ten likert items for Old Spice and the respondents’ movement pattern and gender. The chart below has the results for the MANOVA test:
In this MANOVA analysis there are a few results that stand out as significant in furthering of understanding of the dependent and independent variables. For men on the prompt “It is not important to me to use Old Spice” there was a pronounced difference in mean scores for up-movers relative to down-movers with same respondents in the middle of the two cross sections. The up-movers had a mean score of 4.1 which was .9 higher than the down-movers. The difference in mean between the cross sections of up-moving males and down moving males was beyond one standard deviation for both groups. Therefore, one would expect to see very few if any respondents that defied this result.
|
Wilks’
Lambda |
F-Ratio |
Degrees of Freedom |
Error
Degrees of Freedom |
Gender |
0.41* |
7.22* |
10 |
51 |
Change Score for Old Spice |
0.71 |
0.97 |
20 |
102 |
Change Score for Old Spice by Gender |
0.83 |
0.48 |
20 |
102 |
*P≤.15
For women the likert item which has the prompt “Using Old Spice fits my lifestyle” had higher mean scores for up-movers relative to down-movers beyond one standard deviation. This is the only likert item for women where the result between up and down movers and their mean score is beyond one standard deviation difference. For the majority of likert items, there is little differentiation between women and their move score relative to their likert score. In seven out of the ten likert items the mean scores are within .4 of each other for up-moving women and down moving women.
Similar to the result for the ANOVA, the only significant Wilkes Lambda and F-Ratio is the ratio between gender and the ten likert items. We could expect to see similar ratios if we did this study with the general population. The relationship between the change score for Old Spice pre to post advertisement and the ten likert items is close to statistically significant, as it has a reasonably high F-Ratio of 0.97. However, it did not meet our criteria of being significant at a .15 level. The F-ratio is also not statistically significant for the respondent’s gender and change score relative to the ten likert items. The lack of statistical significance for the two relationships means that the analysis we made for the MANOVA test for our sample population cannot be applied to the general population.
Back to the top
Factor Analysis
I performed a factor analysis for each brand of body wash to place related brand attributes into independent groups, for the calculation of a Brand Attitude Score. I used the 10 Likert items as variables. I then performed a paired t-test to find the potential statistical significance for the differences of means between brand attitude scores. The study consists of 66 voluntary respondents.
Communalities |
|
Old Spice |
Lever 2000 |
Dial |
Likert Item |
h-squared |
h-squared |
h-squared |
___ is a good brand of body wash |
0.7 |
0.7 |
0.8 |
I like the smell of ___ |
0.6 |
0.7 |
0.7 |
___is too expensive |
0.8 |
0.8 |
0.6 |
___ does a better job of getting me clean |
0.8 |
0.7 |
0.7 |
___ helps the health of my skin |
0.5 |
0.6 |
0.8 |
___ is not better than other brands |
0.6 |
0.4 |
0.6 |
Using ___ fits my lifestyle |
0.7 |
0.6 |
0.6 |
The opposite sex likes the smell of ___ |
0.6 |
0.7 |
0.8 |
It is not important to me to use ___ |
0.3 |
0.3 |
0.6 |
I prefer another body wash over ___ |
0.7 |
0.7 |
0.6 |
Communality is the percentage of variance in the original variable that is explained by the factors within the factor analysis. Above you can see that the “___ is a good brand of body wash” and “___ does a better job of getting me clean” Likert items have the most explained variance. This likely could be explained by the fact that those were two of the Likert items that spoke most directly to brand preference. One would expect to see the least variance in the prompts that relate most directly to brand preference.
Total Variance Explained |
|
Old Spice |
Lever 2000 |
Dial |
Component |
Eigenvalue |
% of variance |
Cumulative% |
Eigenvalue |
% of variance |
Cumulative% |
Eigenvalue |
% of variance |
Cumulative% |
I |
4.1 |
41.0 |
41.0 |
3.7 |
37.0 |
37.0 |
3.1 |
30.5 |
30.5 |
II |
1.1 |
11.4 |
52.4 |
1.4 |
14.0 |
50.9 |
1.4 |
14.3 |
44.9 |
III |
1.0 |
10.2 |
62.7 |
1.2 |
11.5 |
62.4 |
1.2 |
12.0 |
56.8 |
IV |
0.9 |
9.3 |
72.0 |
0.9 |
8.7 |
71.2 |
1.0 |
10.5 |
67.4 |
V |
0.9 |
9.0 |
81.0 |
0.9 |
8.6 |
79.8 |
0.9 |
9.1 |
76.5 |
VI |
0.6 |
5.8 |
86.8 |
0.6 |
6.5 |
86.2 |
0.8 |
7.6 |
84.1 |
VII |
0.5 |
4.9 |
91.7 |
0.5 |
4.8 |
91.0 |
0.6 |
6.0 |
90.1 |
VIII |
0.4 |
3.8 |
95.5 |
0.4 |
3.7 |
94.7 |
0.4 |
3.9 |
94.0 |
IX |
0.2 |
2.4 |
98.0 |
0.3 |
3.0 |
97.6 |
0.3 |
3.4 |
97.4 |
X |
0.2 |
2.0 |
100 |
0.2 |
2.3 |
100 |
0.3 |
2.5 |
100.0 |
Old Spice had three factors that were significant in that they had an Eigenvalue over one. A factor with an Eigenvalue less than one means that it explains less than any one of the variables individually which thereby renders it insignificant. 62.7 percent of the variance for Old Spice is explained by the three factors. 37.3 percent of Old Spice’s variance is unexplained variance.
Lever 2000 had three significant factors that explained 62.4 percent of the variance. This result is very close to the result of Old Spice indicating that the two brand had similar response patterns. Dial had four significant factors that explained 67.4 percent of its variance. 32.8 percent of Dial’s variance is unexplained variance.
Old Spice Factor Matrix |
|
I |
II |
III |
___ is a good brand of body wash |
0.7 |
0.2 |
0.0 |
I like the smell of ___ |
0.7 |
-0.1 |
0.3 |
___is too expensive |
0.0 |
0.1 |
0.9 |
___ does a better job of getting me clean |
0.0 |
0.9 |
0.2 |
___ helps the health of my skin |
0.7 |
0.0 |
-0.2 |
___ is not better than other brands |
0.3 |
0.7 |
-0.1 |
Using ___ fits my lifestyle |
0.7 |
0.4 |
0.1 |
The opposite sex likes the smell of ___ |
0.6 |
0.4 |
0.2 |
It is not important to me to use ___ |
-0.4 |
-0.1 |
-0.3 |
I prefer another body wash over ___ |
-0.7 |
-0.1 |
0.2 |
Lever 2000 Factor Matrix |
|
I |
II |
III |
___ is a good brand of body wash |
0.8 |
0.1 |
0.1 |
I like the smell of ___ |
0.8 |
0.2 |
-0.1 |
___is too expensive |
-0.3 |
0.1 |
0.9 |
___ does a better job of getting me clean |
0.1 |
0.8 |
0.0 |
___ helps the health of my skin |
0.3 |
0.6 |
0.3 |
___ is not better than other brands |
0.5 |
0.4 |
-0.1 |
Using ___ fits my lifestyle |
0.7 |
0.4 |
0.1 |
The opposite sex likes the smell of ___ |
0.6 |
0.4 |
0.2 |
It is not important to me to use ___ |
-0.4 |
-0.1 |
-0.3 |
I prefer another body wash over ___ |
-0.7 |
-0.1 |
0.2 |
Dial Factor Matrix |
|
|
I |
II |
III |
III |
___ is a good brand of body wash |
0.6 |
0.2 |
0.0 |
0.5 |
I like the smell of ___ |
0.7 |
0.0 |
-0.3 |
0.3 |
___is too expensive |
0.3 |
0.5 |
0.4 |
-0.2 |
___ does a better job of getting me clean |
0.7 |
-0.2 |
0.1 |
-0.4 |
___ helps the health of my skin |
0.7 |
0.1 |
0.4 |
-0.3 |
___ is not better than other brands |
0.5 |
-0.4 |
0.3 |
0.3 |
Using ___ fits my lifestyle |
0.7 |
0.1 |
-0.2 |
0.0 |
The opposite sex likes the smell of ___ |
0.4 |
0.4 |
-0.6 |
-0.2 |
It is not important to me to use ___ |
-0.2 |
0.4 |
0.5 |
0.3 |
I prefer another body wash over ___ |
-0.2 |
0.7 |
-0.2 |
0.0 |
Using the factor matrix one can see that Old Spice had only one ambiguous Likert item, which was the prompt “It is not important to me to use ___.” Seven of the ten Likert items for Old Spice were in factor one following the lead of the brand preference question. This trend should be expected as a respondent’s answer to the brand preference question is likely a strong determinant to future questions.
The factor matrix for Lever 2000 had the same ambiguous Likert item as Old Spice. Similar to the result the communalities, the congruencency in the results between Old Spice and Lever indicate a similarity in responses. Lever only had six of the ten Likert items follow the lead of the initial brand preference question. Dial was the only brand that had four factor matrices and did not have any ambiguous Likert items. Dial, like Lever 2000, had six of the ten Likert items follow the initial brand question in the first factor.
Attitude Scale (N=66) |
Brand |
Mean |
Standard Deviation |
|
Old Spice |
3.3 |
0.5 |
|
Lever 2000 |
3.2 |
0.5 |
|
Dial |
3.1 |
0.5 |
|
Attitude Scale (inferential statistics) |
Paired Samples |
t-ratio |
Old Spice vs. Lever 2000 |
1.31* |
Old Spice vs. Dial |
2.04* |
Lever 2000 vs. Dial |
1.00 |
The closeness in means on the attitude scale reflects the parity in responses to the survey for the three different brands. Each share a .5 standard deviation meaning that on a scatterplot one would see a lot overlap in data points. The correlation between the attitude scale for Old Spice and Lever 2000 is statistically significant with a t-ratio of 1.31. This means that this relationship could be expected to be similar in the general population. The relationship between Old Spice and Dial is also statistically with a high t-ratio of 2.04. Unfortunately, the relationship between Lever 2000 and Dial was just short of being statistically significant at .15 level.