The data were subjected to two analyses of variance – one univariate, one multivariate. The independent categorical variables for these tests were mover group (up-, down- or non-) and heavy usage. Heavy users are those respondents who have taken over-the-counter pain relief medication at least four times in the last 30 days. In the first test, the dependent variable is the respondent’s Likert score in response to the prompt “Tylenol provides effective pain relief.” In the second test, the dependent variables are the Likert scores for each of the brand attribute questions on the survey.
Table 1: Mean scores for different groups on "Tylenol provides effective pain relief"
|
Heavy users |
Non-heavy users |
Up-movers |
13 |
13 |
Non-movers |
18 |
40 |
Down-movers |
15 |
27 |
Table 2: Tests of between-groups effects
|
Sum of squares |
Degrees of freedom |
Mean square |
F |
Between groups variation: |
2.4 |
1 |
2.4 |
4.66* |
Between groups variation: |
0.9 |
2 |
0.5 |
0.89 |
Between groups variation: |
0.8 |
2 |
0.4 |
0.75 |
Within groups variation |
61.3 |
120 |
0.5 |
|
|
|
|
|
* p ≤ 0.15 |
In 85 samples out of 100 samples drawn from the same population as this sample, it would be expected that the difference in the mean Likert scores on “Tylenol provides effective pain relief” between the heavy users group and the non-heavy users group would be about what it is here. The results of this test can be projected to the population. The variation between the up-, down- and non-mover groups and the variation of the interaction of heavy usage and mover group are not statistically significant and cannot be projected to the population.
Table 3: Multivariate analysis of variance
|
Up-movers |
Non-movers |
Down-movers |
|||
Heavy |
Non-heavy |
Heavy |
Non-heavy |
Heavy |
Non-heavy |
|
Sample size |
13 |
13 |
18 |
40 |
15 |
27 |
Éis a good brand |
4.2 (±0.4) |
3.7 (±1.0) |
3.9 (±0.6) |
4.1 (±0.7) |
4.1 (±0.5) |
4.2 (±0.7) |
Éworks fast |
3.7 (±0.8) |
3.3 (±1.0) |
3.6 (±0.7) |
3.8 (±0.8) |
3.8 (±0.9) |
3.6 (±0.8) |
Éprevents heart attacks |
3.1 (±0.6) |
2.4 (±0.7) |
2.6 (±0.8) |
2.6 (±0.8) |
2.8 (±0.9) |
2.7 (±0.8) |
Éis safer |
3.2 (±0.9) |
3.3 (±0.6) |
3.1 (±0.3) |
3.0 (±0.8) |
3.1 (±0.8) |
3.1 (±0.8) |
Éis trustworthy |
3.9 (±0.8) |
3.5 (±0.8) |
3.6 (±0.7) |
4.0 (±0.9) |
3.9 (±1.0) |
3.9 (±0.9) |
Éwould be prescribed by my doctor |
3.6 (±0.7) |
3.5 (±0.8) |
3.8 (±0.6) |
3.7 (±0.7) |
3.9 (±0.7) |
3.5 (±0.9) |
Éis easy to use |
4.0 (±0.4) |
4.2 (±0.8) |
3.9 (±0.8) |
4.0 (±0.8) |
4.0 (±1.2) |
4.0 (±0.8) |
Édoesn’t have too many side effects |
3.8 (±0.7) |
3.5 (±0.8) |
3.5 (±0.7) |
3.6 (±0.8) |
3.5 (±1.0) |
3.9 (±0.7) |
Éis made by a company that cares about my health |
3.4 (±0.5) |
3.3 (±0.9) |
3.1 (±0.4) |
3.1 (±0.9) |
3.5 (±0.7) |
3.1 (±0.8) |
Éprovides effective pain relief |
4.1 (±0.3) |
3.5 (±1.1) |
3.9 (±0.6) |
3.8 (±0.9) |
4.1 (±0.5) |
3.9 (±0.6) |
Table 4: Multivariate tests
|
Wilks’s Lambda |
F |
Between-groups degrees of freedom |
Within-groups degrees of freedom |
Heavy usage |
0.89 |
0.69 |
20 |
222 |
Mover group |
0.89 |
1.36 |
10 |
111 |
Mover group × heavy usage |
0.82 |
1.14 |
20 |
222 |
|
|
|
|
* p ≤ 0.15 |
The results of the multivariate analysis of variance cannot be projected to the population. In 85 samples out of 100 samples drawn from the same population as this sample, we could not necessarily expect to get the same results.
