Originally Posted by freaknastee
--------------------
Ok after writing all of that and reflecting on it, we do need a bigger sample >.<
I'll rock more hit rating and take another WWS sample and post an update...
|
Yes, your sample size is way too low to distinguish 2.1% vs 1.2%. This is especially
true as this is a sample size over different bosses that may or may not have frost
resistance mucking things up.
There are statistical techniques that will tell you how big a sample you need to
have a 99% confidence rating in the result and the deviation you can expect for
a given sample size. I was going to get all geeky and work it out but I'm rusty
by about 2 years on it and my license for my fancy statistics tool expired so I'm
afraid of giving out misinformation.
So the following is conservative, based on hazy memories of trying to distinguish
similar variances in quality projects:
For a good test, get a tank and healer, find a L70 raid boss they can tank pretty
much forever and spam frost bolt 1s. Get a sample size of at least 500 and
that's probably good enough to distinguish a 1% shift with reasonable confidence.
If you're trying to distinguish tenths of percents, you'll need more like 10k.