Youtube videos to watch:

How to add the add-ins:
http://www.youtube.com/watch?v=sknFUKvR2Fs

For those using Macs, some options like the Data Analysis add-in are
http://www.analystsoft.com/en/products/statplusmacle/
and the somewhat older
http://www.macupdate.com/app/mac/27225/statplus
I wouldn't pay for the full version on those, though--I'd just use the free demo.

I think there's a similar add-in for OpenOffice. 


Then, watch this video about doing a multivariate regression:
http://www.youtube.com/watch?v=x5oxy5io7BY

Then download the file 
http://mapleta.emich.edu/aross15/coursepack3419/multivar-regress-examples-v2.xls

and try it yourself (feel free to re-watch the video as necessary). Note that the file you're downloading has the columns shifted 2 to the right, so what was in Column B in the video is now in column D, etc.

Notice the awesome color coding: the Yellow column is your Y variable.

Examine the residual plots. Do you see any patterns?

After successfully doing the regression, use the results to predict the test score of a district with a % free lunch of 20% (might need to use 0.20 instead of 20?), an average teacher salary of $50,000 and a student/teacher ratio of 15. You should get a prediction of 1335.55703.

Then, try doing the regression again but also include column C, the district # (so now your x variables are C2:F535). Look at the R^2 value in the regression sheet (cell B5). Is it higher or lower than it used to be?

Then, try doing the regression again but also include column B, which is just random. Now your x variables are B2:F535. 
Look at the R^2 value in the regression sheet (cell B5). Is it higher or lower than it used to be?

Then try it again also using column A along with the other columns (so now your x variables are A2:F535).
Look at the R^2 value in the regression sheet (cell B5). Is it higher or lower than it used to be?

What can you conclude about adding variables to your model, even if those variables are clearly junk and shouldn't make the model any better?

Okay, now go to the next tab (HeatIndex) and do a regression to try to predict HeatIndex as a linear function of temperature and humidity. Examine the residual plots. Do you see any patterns?

Feel free to send me emails as you progress along this path. Attach your current sheet.

You can also watch other screencasts at
http://cameron.econ.ucdavis.edu/excel/excel.html


Prof. Ross

Here's something we won't need as a class, but perhaps some projects might want it:
How to do multivariable regression without the add-in:
http://www.youtube.com/watch?v=x7f0P_6rtpw

Later in the semester, we'll do this (no need to watch it now): using Solver
http://www.youtube.com/watch?v=I3pckP_8T-k