Sunday, December 9, 2007

Principal Component

Wow... it is 12:22 AM Monday, December 10th. Finals start next Monday, which means that this is the final week of class! Fortunately, Thursday is the last official day of class, making Friday a "reading day." Hooray for reading days. My Stat 336 (6.0 credit hour class-of-death) class is held MWF from 9:00-11:00 AM, which means that Wednesday will be my very last time in that class! Hallelujah! I've spent many hours today (Sunday, actually) trying to do an analysis of a huge set of data. We were given 11 pieces of information on 196 subjects, but we were never told what any of those measurements actually represented. Just a bunch of numbers put under the categories, "x1", "x2"... "x10", and "y." We were supposed to be able to create an optimal model for predicting "y" given new measurements "x1" through "x10." Well, there are about a million different things one can do to analysis data and create a model, but we haven't learned quite enough to successfully (and confidently) overcome problems such as multicollinearity, influential observations and outliers. Sure, we've learned how to DIAGNOSE the problems, but our knowledge of remedial measures are slim.

Why do I bring any of this up? Well, I just finished my own 13 page "analysis" ten minutes ago and Suzanna is asleep in the other room. My mind would never let me relax enough to go to sleep without at least getting some closure with this project. I think I'll be able to relax better now that I've kind of talked over the things that have been occupying my attention for the last who-knows-how-many hours. Too bad I didn't have a blog this summer when I finished reading Harry Potter #7 at 3:00 AM. I took a LONG time falling asleep that night (morning?) after finishing. The ending was riveting and oh-so appropriate. I felt sad, thinking that my adventures with Ron and Hermione and Harry had come to an end.

Anyway, I wish there was a way to post attachments. If I could figure it out (if possible), then I would post a copy of my nasty analysis so anybody interested could look to see all the fun code I had to do. Maybe as a teaser, I'll include a little here...

\begin{comment}
<>=
plot(2:9,rs$cp,xlab="No. of Parameters",ylab="Cp Statistic",main="Cp Statistic Plot")
abline(0,1)
@
\end{comment}

The Cp statistic graph in Figure 3 shows that the optimal number of parameters in the model is 7. Thus, we \emph{will} use a model without those three variables above mentioned.
\begin{figure}[tbh]
\begin{center}
\includegraphics[width=3in]{cpstat}
\caption{Graph of the Cp statistic.}
\end{center}
\end{figure}

It doesn't look as cool as plain text. I have to type all the stuff in a wordpad document and then run it through a program called "R" (a statistical software program) using a function called, "Sweave." Then I use LaTeX (a typesetting program) to turn my .tex document into a fancy PDF file. In the end it looks great and professional. Having to figure it all out at the beggining of the semester was a headache. I feel like I've learned a ton this semester, and not all of it being pure statistics! Hah!