Comments | b_canadensis: Dear weather, why are you so cold?

b_canadensis

Dear weather, why are you so cold?

Mar 12, 2009 17:07

Seriously, it is freezing out today. It was cold, windy and damp yesterday, but today it's freezing. And this is made more irritating by the fact that only a few days ago it was actually relatively nice out (ie. the temperature was above the freezing mark.) Today it feels like it's the dead of winter again. It's not even that I mind the cold, ( Read more... )

weather, theatre, stupid statistics

Comments 7

c_canadensis March 13 2009, 01:52:59 UTC

Oh, look what I've done. Yes, they're supposed to be robust to violations of normality but *technically* if you talk to a hard-core stats person they're likely to tell you that the results aren't trustworthy if the residuals are non-normal... of course, many of us ignore those people.

Were your data actually continuous or was the range really restricted or something?

Agreed on the "die in a fire" bit, could a branch of mathematics actually die.

b_canadensis March 16 2009, 06:18:09 UTC

Yes, right, I feel that this can totally be blamed on you. Good, I now have a course of action.

Both my residuals and my data are bimodal. Woo. Hoo. The data is continuous though, and the range isn't restricted, and there aren't that many outliers.

We'll see what my supervisor thinks tomorrow...

How could statistics be killed.... hmmmm... new project...

A thought on your stats c_canadensis March 16 2009, 21:21:51 UTC

... from one who has no idea what your analysis is trying to show and has never run AMOVAs but assumes they're somewhat similar to ANOVAs:
Is there some extra variable that might explain why it's bimodal, so you could either add that to the model, or split the data by that variable and analyse each group separately?

Sure your advisor came up with the correct solution anyway, just curious.

Re: A thought on your stats b_canadensis March 16 2009, 22:11:11 UTC

I meant ANOVAs actually. AMOVAs don't usually so much have that problem since they're usually on sequence data (which aren't so much normal as "Oh, that base pair's different..."). But, after a visual analysis of the residuals, we believe they are normal enough. Also, I'm trying to do a box cox transformation, if JMP will let me that is. If it's not the stats, it's the stats programs...

Re: A thought on your stats c_canadensis March 17 2009, 19:55:45 UTC

Ooh, you have JMP? How are you liking it? I want to try it because it's the user friendly version of SAS (which I've been learning), and amazingly has a Mac version.

Re: A thought on your stats b_canadensis March 18 2009, 14:29:06 UTC

It's alright. It's fairly user-friendly. It's actually what I used in undergrad too. I have the student edition, which annoyingly doesn't have all of the features on it (althought he help function thinks it does, stupid deluded help menu). And we can't get the Mac version to load on any of the computers.

Thread 5