Exploratory Study -- ordinal data questions

JB12

New Member
#1
Hello all, I’m a first time poster but have gained much insight reading through past threads.

I’m running an exploratory study, containing a few different assessments, but I’m most concerned with the following:

All participants will experience 15 scenario-based questions, where they are to rank (1=least important – 5=most important) 5 characteristics for each question. Although these characteristics are disguised, they are constant for each question. The questions will be grouped into three different levels (5 questions per level), determined by threat (high threat scenario/med threat/low). Again, all participants will complete the 5 questions within each level (these levels will be disguised and counterbalanceed).

First, I’m interested in seeing if demographic data such as age/experience will affect how people prioritize these options. I could even take it a step further and see how demographic data affects how people prioritize data in each threat level. I’m not exactly sure the most appropriate way to do this. However, I’m thinking a Wilcoxon-Mann or Whitney Test will do the trick. Thoughts?

Secondly, I’m attempting to see if threat level affects how participants (as a whole) rank these characteristics. I was thinking about using Friedman’s Test, but I’m not sure if that is what I want.

Any thoughts would be greatly appreciated. I can clarify the study if needed.

Thanks,
Joe
 

noetsi

No cake for spunky
#2
If you are adding the levels of multiple ordinal questions to get a single measure then it is usually accepted that you can treat this as interval data. For instance if you measure the response to some phenomenon by adding five ordinal questions (thus getting a response from 5-25) you can probably treat this as interval like (the more levels you get the more this is true - so that if you had ten questions collapsed to get a measure from 10-50 this would be more interval like).

If your measurements are interval like you can use methods such as regression or ANOVA that are commonly preferred to non-parametrics you answered.
 

JB12

New Member
#3
This is great. Thank you for the quick reply. I'm unfamiliar with ranked/ordinal data, so I didn't know you could treat it as interval data like you mentioned.

Thank you very much!
 

noetsi

No cake for spunky
#4
I will note in passing that while this is commonly done, for instance in education research, you almost certainly could find individuals who will disagree. Few things in statistics seem to be in as much dispute as the use of likert data.