Ranking question

puffin24

I have survey data that I need to analyse. Respondents have ranked the impact of something on a scale from 1-5 (1= mild, 5= v. severe etc) I am classifying means from 1-2 as mild, 2-3 moderate, 3-4 severe and 5 as v. severe but obviously due to the design of the survey a mean score of 5 isnt possible, so the result gets dragged down; eg for one of the cases half of the respondents selected 5, but the mean will only regard this effect as severe at most. Should I weight those who select 5, or should I have more flexible cut off points, e.g a mean of below 4.5=severe and above this= v.severe? I hope this is the right section for this question and that I havent made the question needlessly complicated!

Yakuza

Ideally, each label would only have one number associated with it, have you let them answer with decimal numbers (as in 3.8), or only choose 1 or 2 or 3 or 4 or 5?
Why not have 1 = v mild, 2 = mild, 3 = moderate, 4 = severe and 5 = v. severe?
You're going to tie yourself in knots having 4 categories but 5 answers.

If they could give a decimal answer (i.e from 1.0 to 5.0) then you could assign different categories / buckets to different ranges, eg:
1-1.5 v. mild
1.6-2.5 mild
2.6-3.5 moderate
3.6-4.5 severe
4.6-5.0 v. severe

puffin24

Thanks for your suggestion. The respondents did answer in whole numbers but I would like to report the mean; e.g. 4.5 for property value impacts etc. I have decided to include the standard deviation because that will give an indication of the range, to show that some did answer 5 for some of the impacts.