In your analysis you talk about using better measures than the mean and standard deviation. I would go further and argue that the mean and standard deviation, and median and percentiles are not valid for the way they have processes the data from the chandler fatigue scale. Basically they say lets treat it as a likert scale and add up all the values and use this as a measure. That is simply wrong!
The chandler fatigue scale is a multi factor scale (including influances of physical fatigue and mental fatigue - look at the principle component analysis quoted in their original paper). A likert scale requires all the questions to relate to one underlying thing or attitude. You should show there is a high degree of correlation between the answers of all the questions. What they have created is a scale that is not necessarily monotonic - that is an increase the things you are measuring will not necessarily lead to an increase in the measure they are using. They could talk about the mode but that wouldn't be very useful.
Even if they had a single thing they were measuring they need to worry about whether it is an ordinal or interval scale. That is is the scale linear - does the gap between 1 and 2 represent the same gap as between 4 and 5. This is complicated by two factors, firstly the scale on which questions are answered. People answering the scale need to understand it as even and the scale needs to represent the full extremes of the possible answers. If we see bunching at the end of a scale then the end points probably represent a much bigger interval than the middle points.
The second factor is when you add up all the results of different questions, since there are no units it is hard to say that the resulting sum evenly covers the overall scale. I beleive psyciatrists tend to say but our intervals are of the same order of magnatide hence its ok.
If the scale is really an interval scale then you can use the mean and std and do things like regression on the results however if it is not you cann't.
As long as the results are ordered even if not evenly then you have an ordinal scale which you can quote the median and percentiles for. But the mean and std make no sense.
I would argue that if their questionaires only measure one (unseperable) concept then they have an ordinal scale. They may be able to argue that they have an interval scale but it is up to them to demonstrate that they have the correct properties. Where they have things like the chandler fatigue scale then there analysis is just wrong as its measuring some unknown function of two or more different things.
The chandler fatigue scale is a multi factor scale (including influances of physical fatigue and mental fatigue - look at the principle component analysis quoted in their original paper). A likert scale requires all the questions to relate to one underlying thing or attitude. You should show there is a high degree of correlation between the answers of all the questions. What they have created is a scale that is not necessarily monotonic - that is an increase the things you are measuring will not necessarily lead to an increase in the measure they are using. They could talk about the mode but that wouldn't be very useful.
Even if they had a single thing they were measuring they need to worry about whether it is an ordinal or interval scale. That is is the scale linear - does the gap between 1 and 2 represent the same gap as between 4 and 5. This is complicated by two factors, firstly the scale on which questions are answered. People answering the scale need to understand it as even and the scale needs to represent the full extremes of the possible answers. If we see bunching at the end of a scale then the end points probably represent a much bigger interval than the middle points.
The second factor is when you add up all the results of different questions, since there are no units it is hard to say that the resulting sum evenly covers the overall scale. I beleive psyciatrists tend to say but our intervals are of the same order of magnatide hence its ok.
If the scale is really an interval scale then you can use the mean and std and do things like regression on the results however if it is not you cann't.
As long as the results are ordered even if not evenly then you have an ordinal scale which you can quote the median and percentiles for. But the mean and std make no sense.
I would argue that if their questionaires only measure one (unseperable) concept then they have an ordinal scale. They may be able to argue that they have an interval scale but it is up to them to demonstrate that they have the correct properties. Where they have things like the chandler fatigue scale then there analysis is just wrong as its measuring some unknown function of two or more different things.