Removing extremes from a statistical range

[Terryfic10]

Hi

I have a range of scores given by visitors using a scale from A to F were A is Excellent and F is Poor. These are weighed to give an overall performance score, so lower scores (E or F) can have a dramatic effect on the overall performance score.

The scores are nearly always A or B, but in one recent survey 24 out of 25 rated the experience as A/B and oneperson rated it as F.

My question is this – is it statistically valid to ignorethis grade when calculating the overall performance scale? I seem to remember readingsomething once about ignoring extremes of a range when calculating the mean –or did I imagine it?

Regards

Terry

 

[Ishuda]

Hi

I have a range of scores given by visitors using a scale from A to F were A is Excellent and F is Poor. These are weighed to give an overall performance score, so lower scores (E or F) can have a dramatic effect on the overall performance score.

The scores are nearly always A or B, but in one recent survey 24 out of 25 rated the experience as A/B and oneperson rated it as F.

My question is this – is it statistically valid to ignorethis grade when calculating the overall performance scale? I seem to remember readingsomething once about ignoring extremes of a range when calculating the mean –or did I imagine it?

Regards

Terry

What you are talking about is statistical outliers, see
http://en.wikipedia.org/wiki/Outlier
for example. As the article says, it depends on the reason you have an outlier (outliers) in your data. Certainly, if you strongly suspect that the rating was mistakenly marked, either by accident or on purpose, you should discard it. But only you can decide that. If you do decide to keep it, then any model you use (such as a single mode model) might have to be re-examined for appropriateness.

 

[Terryfic10]

Outliers in statistics

Thanks - the article was very useful. I just could not remember the correct term.