The Truth About Cats and Dogs in Black and White

Truth is just one person's reality. Here is mine in black and white (maybe some shades of gray). This blog has little to do with Cats or Dogs - just humourous sarcastic antics about my life or occasionally, someone else's. You know, intercepting volleyballs with my face, egg dropping, etc. The truth has seen some changes and in fluidity with change expect to see more artistic expressions posting up - so give your two cents worth!

Wednesday, February 08, 2006

What do you call a Dino that loves Stats?


A statasaurus of course. And obviously they are only in it for the goats. First, I must thank the cartoonist. I totally rogued this off the internet. I spent part of my day in a meeting with my statistician. It transpired as well as expected - it is stats after all. If only my stats trouble revolved around goats and probabilities.

The meeting began with the usual interrogation.

Q: Did you standardise your data?
A: Standardise? Doesn't the PCA do that for you? (To self: That would be a no)

Q. You tested for univariate outliers. Did you test for multivariate outliers?
A: (To self: I did?) Ah...no? (wait for it wait for it) what exactly is the difference? (Bloody hell! Is he speaking english?)

Q. Did you find any univariate outliers?
A. Outloud: Uhm...honestly I have no idea.

Prognosis:
  • Recode dependent variable in 2 data files.
  • Standardise all ordinal and interval data ~ 200 variables.
  • Run linear regression to find Mahalanobis distances for each of 4 data sets and find Chi square values for each sets' df.
  • Rerun Principal Component Analysis (PCA) using the standardised data set.
  • Rerun PCA by selecting only those cases whose z scores are less than the Chi square (thus eliminating multivariate outliers.
  • Compare the KMO and determinate values for both PCA's to determine if the outliers influence the components.
  • Check the anti-image for something which I can't remember.
  • Check the rotated component matrix for redundancy amongst variables and check for clearly definated components.
  • Schedule another meeting with statistician.

Perfectly straightforward, yes? Well such is my day tomorrow. No wonder the bloody dinosaurs went extinct! Oh, what I would give for a goat!

Unknown 8:05 p.m.
Comments...

Sounds like a fun time.

Nothing like linear regression to get the heart pumping.

Oh, and I hope you meant one of those little goats because they are quite cute.

posted by Blogger justacoolcat : 8:48 p.m.  

I will admit that even I, who has a degree in math, am confuzzled by all the stats talk. Obviously I didn't go far enough into stats (being a mathematical modelling gal myself). Or maybe I just went far enough.

Good luck.

posted by Blogger Laurie Roche : 12:28 p.m.  

Are you still working?

Have some fun with FFF.


BTW,Laurie, you are right on. The real world needs and loves mathmatic modeling way more than stats, well atleast I do.

posted by Blogger justacoolcat : 11:55 p.m.  


Post a Comment



<< Home