Consumer Preference Survey:
The Story of Three Pasta Sauces.
Table of Contents
Frequency Table: Change Scores
Frequency Table: Brand Preference
Paired t-test of Female Subjects
Two pairings are statistically significant atp≤ .15 (Ragu vs. Newman’s Own and Bertolli vs. Newman’s Own). The results from these two pairings can be generalized to the rest of the population with a particular degree of certainty. Specifically, in 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find mean score differences between Ragu and Newman’s Own and also Bertolli and Newman’s Own about the same as shown in this sample.
It was found that respondents that moved up and those that moved down had significantly different Brand Index Scores and Ad Index Scores. In other words, in 85 or more samples out of every 100 samples drawn from the same population as this sample we would expect to find the Brand Index Score and Ad Index Score about the same for respondents that moved up and also those that moved down and that these results can be projected to the total population.
The chi-squared test revealed a significant difference at the p≤ .15level meaning that in 85 or more samples out of every 100 samples drawn from the same population as this sample we would expect to find the frequencies for up movers, down movers, and same score respondents about the same as above.
Frequency Table: Change Scores
This is a frequency table showing simply how many of the 72 respondents moved up or down or stayed the same in their pre- to post-test scores. These are descriptive statistics of the sample and so no statistical analysis was conducted, hence there are no inferential results to report.
Frequency Table: Brand Preference
This is a frequency table showing how many of the 72 online respondents preferred Newman’s Own over Ragu or vice-versa according to each brand’s particular Brand Index Score for each respondent. In this particular comparison, there were no respondents who liked both equally, so the results are split into 53 liking Newman’s Own more and 19 liking Ragu more. These are descriptive statistics of the sample and so no statistical analysis was conducted, hence there are no inferential results to report.
This is a correlation test to determine the strength of the relationship between the Brand Index Scores of Ragu and Bertolli. A correlation coefficient of -0.6 indicates a high coefficient between the two scores. Specifically, there is a high likelihood that, within a respondent, as one Brand Index Score goes up, the other goes down.
Paired sample t-test of female subjects
It was ascertained that females did significantly prefer Newman’s Own over Ragu. In 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find mean score differences between Ragu and Newman’s Own about the same as shown in this sample.
Ragu
The first table shows statistics describing the relationship of the independent variables (the Likert items) to the dependent variable (the move score). It was concluded that the F-ratio is significant at thep≤ .15 level. This indicates that in 85 or more out of every 100 samples drawn from the same population as this sample, the coefficient of multiple determinants would be similar to those in this sample. The second table shows statistics for the constant and also each Likert item individually. Those scores highlighted in red are the most important predictors of the change score, whether it be ‘up’, ‘down’, or ‘same’. In 85 or more out of every 100 samples drawn from the same population as this sample, the items indicated by an asterisk would be significant indicators of the change score as shown in this sample.
Bertolli
The first table shows statistics describing the relationship of the independent variables (the Likert items) to the dependent variable (the move score). It was concluded that the F-ratio is significant at thep≤ .15 level. This indicates that in 85 or more out of every 100 samples drawn from the same population as this sample, the coefficient of multiple determinants would be similar to those in this sample. The second table shows statistics for the constant and also each Likert item individually. Those scores highlighted in red are the most important predictors of the change score, whether it be ‘up’, ‘down’, or ‘same’. In 85 or more out of every 100 samples drawn from the same population as this sample, the items indicated by an asterisk would be significant indicators of the change score as shown in this sample.
Newman’s Own
The first table shows statistics describing the relationship of the independent variables (the Likert items) to the dependent variable (the move score). It was concluded that the F-ratio is significant at thep≤ .15 level. This indicates that in 85 or more out of every 100 samples drawn from the same population as this sample, the coefficient of multiple determinants would be similar to those in this sample. The second table shows statistics for the constant and also each Likert item individually. Those scores highlighted in red are the most important predictors of the change score, whether it be ‘up’, ‘down’, or ‘same’. In 85 or more out of every 100 samples drawn from the same population as this sample, the items indicated by an asterisk would be significant indicators of the change score as shown in this sample.
Discriminant Analysis
A discriminant analysis was performed to see if the up-movers and down-movers could be separated into statistically significant groups and to see which Likert questions were most influential in the division. The Chi-squared figure shown is an indirect way of testing Wilk’s Lambda and it was found to be significant. In 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find group centroids for up and down-movers about the same as shown in this sample. Also, the confusion matrix has successful group prediction rate of 90.3%. That is to say of the respondents that took the test, 90.3% were correctly predicted to be either up-movers or down-movers. Since the test achieved a t score of 5.89 with critical t at 1.04, this confusion matrix can be projected to the population. Specifically, In 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find a successful prediction rate about the same as found here.
The inferential statistics show that the F-ratio is not significant for any of the independent variables or the interaction effect on the dependent variable. We can not project these results onto the population. Specifically, in 85 or more samples out of 100 samples drawn from the same population as this sample, we could not expect to find categorizations showing relationships between the independent variables and the dependent variable as well as the interaction effect on the dependent variable about the same as shown in this sample.
MANOVA
The inferential statistics, shown in the table below them, are of much greater interest. The Wilks’ Lambda is significant for both independent variables as well as the interaction effect on the 10 dependent variables taken together. Specifically, in 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find categorizations showing relationships between the independent variables and the dependent variable as well as the interaction effect on the dependent variable about the same as shown in this sample. Unlike the ANOVA which was not significant for only one dependent variable, when all 10 questions are considered, the results are generalizable to the population.
All statistics derived directly from a factor analysis are descriptive; no generalizations can be made to the population from them alone. But, given the variables which load onto the same factor as the ‘__ is a good pasta sauce’ Likert, an attitude scale was achieved with mean scores ranging from 1 to 5 with 1 being the most favorable and 5 the least. This is reasonable as ‘__ is a good pasta sauce’ is an evaluative question, so, by definition, all variables that load onto the same factor are also evaluative and contribute to the overall attitude toward the brand. After t-tests were run between the mean scores it can be determined that in 85 or more samples out of 100 samples drawn from the same population as this sample, we would expect to find mean score differences between all combinations of the three brands to be about the same as shown in this sample.
