Hi everyone!
In my survey I have 3 likert scales measuring 3 antecedents of brand equity. My aim is to create a variable representing Brand Equity, somewhay summing up the 3 single likert scales, in order to run a ANOVA analysis between intervention and control group.
Can someone help me please?
Hi Matteo,
Would it be possible to run an ANOVA for each Likert scale separately between the experimental or control group? I assume that because you want to combine them, this may not be what you are looking for. I have provided two solutions that solve the issue, however, come with their own limitations.
- Averaging the 3 Likert scale responses (for each participant) and then running an ANOVA on the average.
- Sum all 3 Likert scale data (depending on how they are formatted) to give a single value.
Would be good to see how you get along.
Pat
Patrick Swain thank you for your help!
Anyway, it is possible to run ANOVA on the three separate scales, and I have already done it to verify my first 3 hypothesis. The last hypothesis, however, was concerning the effects on overall brand equity, that is why I needed a way to do it.
Did I get what you were suggesting?
Among the two alternatives you proposed, is there one you particularly suggest, or they are equivalent?
Thank you again!
The only difference would be the scale by a ratio of 3 (as there are three questions). As a percentage of the maximum possible score, the methods are identical.
For example:
Assuming there are 3 Likert scales with response values 1-5, and one person reports 3, 4, and 5, for the 3 questions.
The average method would result:
3+4+5 = 12 %sum of Likert value for each question
12/3 = 4 %divide by number of questions (this is the average)
(4/5)*100 = 80% %report as a percentage of max possible score (i.e. 5)
The sum method would result:
3+4+5 = 12 %sum of Likert value for each question
(12/15)*100 = 80% %report as a percentage of max possible score (i.e. 5+5+5 = 15)
It would be good to supplement this 'overall' finding with the individual scale analyses, as these methods may hide important data. For example, say a group of people reported on average, 2, 5, 5 for three questions on a similar topic. The overall score would be very good (i.e. 80%), however, the '2' is an important finding and discussion point.
If your Likert scale questions do not have the same intervals, for example, one had response values of 1-5 whilst another had 1-20, you can convert these into a standard format by using percentages of the maximum score. I.e. 1/5*100 = 20%, would be equal to 20*20% = 4/20. So 1/5 = 4/20.
Although you have already done some stats, it is worth making you aware of the limitations of using parametric tests (e.g. ANOVA) for Likert scale data - for more info see Article How to use survey results
Hope this helps,
Pat
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running