Anyone good at statistics?

[kári]

I have to turn in a report for my Statistics class this afternoon and I seem to have looked over one of the questions...
"The Shapiro-Wilk test for normality uses a correlation that is much like the Pearson correlation. Explain if it is possible to do the same with the Spearman correlation."

I don't t hink I even understand the question fully... "Do the same?" like, build a normality test, that also works for non-normal populations...? Any help appreciated!

Thanks!

[YtseBitsySpider]

"Kent Kent Kent........people can come up with statistics to prove anything...14% of people know that"

[rumborak]

Well, the Spearman correlation is just the ranked version of the Pearson correlation, right? I don't see why, whatever you do with Pearson, you couldn't do with Spearman too.

rumborak

[kári]

Well, the Spearman correlation is just the ranked version of the Pearson correlation, right? I don't see why, whatever you do with Pearson, you couldn't do with Spearman too.

rumborak

Well I turned in the paper and I was pretty sure of my answer, which was that the Pearson correlation checks for a linear relation and thus when done in a Q-Q plot checks for normality, whereas the Spearman correlation can only detect whether there is a relationship between two variables, like when one rises the other does too, etc. So it can't be used to check for normality in a Q-Q plot.

[rumborak]

Interesting, didn't know you can use Pearson for that. But yeah, Spearman I think loses that information because you are converting the values into ranks.

rumborak

[kári]

It's a pretty common test though I think, called the Shapiro-Wilk test.