Simple Data Transformation Question

[KUJustin]

I'm looking for some ideas on how to transform some data I'm working with.

Basically I have a bunch of "average ratings" from some survey data. A number of attributes have been rated by multiple people from 1 to 7. The "average ratings" are generally between 4 and 5 (mean: 4.6, stdev: .7).

I'm looking for a systematic way to "spread out" the data so the values aren't all in the 4 to 5 range. Basically I want to flatten my bell curve.

I've got a handful of ideas I'm just curious what you guys would see as a good way of going about this because I have virtually no formal statistics training.

[tame_deuces]

I'm abit uncertain on what you mean? If you have normal distribution just present it, if you don't then present it that way to and include criticism on that in your methodology section if applicable to the analysis type, population sample or data size you have.

If it is a methodological problem for your statistical test type (if any), then you have to use another test type - you should use a normality test to figure it out.

Some other poster will have to come along if I'm just filling you with nonsense right now, haha.

[Silent A]

This seems too easy ...

1) calculate the z-score of each data point
2) select a new standard deviation (> original SD)
3) use original z-scores with the new SD to obtain "flatter" data set.
4) profit!

Or did you mean something else?

Plus, I can't think of any compelling reason to do this.

[KUJustin]

[ QUOTE ]
This seems too easy ...

1) calculate the z-score of each data point
2) select a new standard deviation (> original SD)
3) use original z-scores with the new SD to obtain "flatter" data set.
4) profit!

Or did you mean something else?

Plus, I can't think of any compelling reason to do this.

[/ QUOTE ]

That was the best idea I'd had so far, was interested to see if I'd either get confirmation of that or another solid idea.

I'm not really sold on us having a compelling reason to do it either, but we're trying to improve some calculations we're doing with that data in comparing it to another set of data, and a transformation of the data was one idea that was suggested.

Thanks for the input.

[stopcensoringme]

Why not draw your curve in pencil next time. Then you make it look how you want it. Better yet, don't even bother with the survey next time. Just make it up from the get go.