Ranking question analysis

I want to understand what is the best to analyze ranking data. We just completed a survey with 132 respondents. They were shown a list of 11 attributes and asked to pick the top 5. What method should be used to identify the top 3 attributes?

Is it okay to do a straight mean if the N is close enough? How do I determine if the N is close enough. For instance, if i do a straight meant, attribute A has a mean of 3.08 (N=114 out of 132) and attribute B has a mean of 2.96 (N=104 out of 132). In this case, since respondents were asked to rank 1 as highest rank, a lower would signify a higher rank.

Of the above 2 attributes, how do I decide which one is better?


TS Contributor
What is the span of the ranking assigned by each respondent (from 1, the highest, to what value?

I hope to have well understood your question and the context of your issue.

May be you could:
- graphically visualize the ranking of the various attributes by using boxplos
- (if 1 is the highest rank) select the attributes with the lowest ranks
- use Kruskal-Wallis test to see if there is a significant difference between attributes as far as ranks are concerned.

Hope this helps.
Feedbacks from other users are welcome.



TS Contributor
please, see the example in the attached .pdf.

I created a fictional example, with 10 respondents and 4 attributes (ranked from 1 to 10).

Boxplots and K-W test provided (analysis from SigmaPlot).

In this case, attribute 1 and 2 are the ones with the lower (highest, to keep with your terminology) rank. The difference in mean rank is not significant between 1 and 2, while both attribute 3 and 4 (that is, those with lower rank in your terminology) significantly differ from 1 and 2.

So, may be you could conclude that 1 and 2 are better than 3 and 4. Attribute 2 is slighly better than 1, but the difference is not significant.

(Please note that, in order to create an example, I did not paid attention to create samples meeting the assumption of K-W!!)

Hope this helps.