Delphi study / 9 point Likert scales

[veena]

We are conducting a Delphi study.

Each statement in our questionnaire is scored on a nine point likert scale ranging from 1(strongly disagree) to 9 (strongly agree).

Previous Delphi studies from our college and department of health have all interpreted 7-9 as agreement, 4-6 as indeterminate, and 1-3 as disagreement.

Can anyone tell me why 7-9 is taken as agreement, when on a nine point scale anything above 5 indicates agreement? I have asked locally, but even though everyone uses this method, nobody knows why.

Many thanks.

 

[JohnM]

Not every response above 5 on a 9-point scale indicates agreement - it depends heavily on the individual's interpretation of the scale.

That's why 4-6 is considered indeterminate - it's "in the middle"

In my opinion, anything above a 5-point scale is a bit of a stretch - I just don't believe that people can really distinguish a difference as narrow as 1 point on a 7 or 9-point scale, and people tend to not use the entire scale (sometimes even at 5 points).

At my place of work I send out an annual 5-point Likert customer satisfaction survey, and I have never, ever received a "1" on any question, ever.

The only exceptions to this might be a poll of subject-matter experts where the subject of the poll itself is their area of expertise.

 

[veena]

Thanks, that is helpful.

This is a poll of experts on their subject matter. Still not sure of the logic of giving them 9 choices if we are going to group the responses anyway.

The other decision that we've had to make is what to accept as consensus agreement for each item. In the literature people have used mean 7-9, median 7-9, upper three quartiles (75% of respondents) 7-9.

I don't want to use the mean because it will be distorted by the outliers. But I'm not sure about median score v 75% of responders. I've gone for 75% as it seems to be the more rigorous of the two, but am worried about excluding too many items, especially as this is an area where there is no 'right' answer.

 

[JohnM]

For consensus I would use the 75% rule or some other % threshold - possibly even higher than 75%. It just seems to make more sense, considering the dictionary definition of "consensus."