# How to show that method X is more reliable than method Y?

I compute a statistic s using two methods, X and Y. In order to measure the reliability of X and Y, I found split half reliability measures (alpha) using a bootstraping technique. So, for each of method X and method Y, I found the split half alpha value N times, where N is the number of bootstraps.

Now, I have N alphas for X and N alphas for Y.

XAlphas = [xa1, xa2, xa3, ..., xan]
YAlphas = [ya1, ya2, ya3, ..., yan]


I'd like to show that X is more significantly more reliable than Y (meaning, the mean alpha for XAlphas should be higher than the mean alpha for YAlphas). What type of statistics test would I use for this? Is this a T Test?

-
I think you should add some detail. Perhaps you don't mean "statistic" in the first sentence. Or maybe you don't mean "reliability". "Split half reliability" and "alpha" (two different things) are usually applied to scales, not statistics. Please tell us what you are trying to do, in non-technical language. –  Peter Flom Aug 25 '12 at 22:00
just think of it as two arrays. i'm trying to show the first array is larger than the second. I do indeed mean statistic, and I do indeed mean split half reliability, which in return gives an alpha if you correlate over multiple subjects –  CodeGuy Aug 25 '12 at 23:07
Just a thought...since alpha is bounded by 0 and 1, would it make sense to use Fisher's Zr transformation along the way to doing a T-test? –  rolando2 Aug 26 '12 at 2:46
I'm not really sure :/ –  CodeGuy Aug 26 '12 at 3:31
I'd suggest re-writing the part of your question that says "found split half reliability measures (alpha) using a bootstraping technique". If you could describe what you are doing here in more straightforward language, and why - what is your actual question here about Y versus X - it would be very helpful. –  Peter Ellis Aug 26 '12 at 9:56