Statistics and averages

[phl666]

Is there a better way of calculating the most popular choice in a set of answers? Normally, you would use the mean (average), but one person choosing a wildly different answer would throw the whole lot. An alternative would be to use the Mode, but if the answers are all different, then this wouldn't work either. Is there another method?

[ugly_bob]

You can use any of the three 'averages'
mean - as you describe
median - the central value when all are listed in order (ie median of {1 1 3 3 5} is 3
mode - the most popular

[phl666]

But a better way would be some combination of mean, median and mode, would it not?

[ugly_bob]

possibly. Something like standard deviation?
where you quote the mean and a representation of how close other figures are to this?
Or i remember something else Chi-squared (?) but can't actually for the life of remember what it represents

[tubeway]

If you quote a mean avrage and then give the standard deviation it will let people know how far from the average a typical figure is likely to be.

[mikeyp]

I may be wrong, but I think Chi-Squared shows the difference between expected and observed results.

As has been suggested the mean and standard deviation are a good way, you can then produce a normal distribution that will allow you to estimate the probability of a certain choice of answer.

Which option depends on the situation. In your case, where you want the most popular choice? A mode measure seems most appropriate.

[ugly_bob]

You might well be right mikey - I just remember it from stats.

[heathfield]

Now you know how it can be that over 50 percent of people are below average height. ;-)

[ugly_bob]

"You can prove anything with statistics. 67% of all people know that" - Viv Reeves

[kwicky]

I always use the 'Normal Distribution' or Bell Curve. Then 99% of the population are covered within 4 standard deviations.

[skilliyay]

Also depends if you a dealing with continous or discreet data. The classic example being "The average family has 2.4 children" When in fact no families do.