AP Stats Project: correlation btwn per capita income and state rank

I just finished my AP stats project last night....anyway, I highly doubt there is going to be that much correlation (from my own experience at least), however good luck even obtaining enough data for worthwhile tests....Not to mention the insane number of confounding variables.

Pick something else.

OKLAHOMA

JENKS- RICHEST COMMUNITY IN OKLAHOMA, CONSISTENTLY 1ST PLACE IN TRACK/XC

US GRANT- POOREST COMMUNITY IN OKLAHOMA. CONSISTENTLY LAST PLACE IN TRACK/XC

YOUR WELCOME

I'm a scatman!

Only thing that's difficult is getting the XC results. Pulling the school district income/other characteristics data is going to be simple. But to reference the Oklahoma example above, there appears to be 522 school districts in that state, so compiling that with income data might be time consuming depending on how the running results are tabulated.

Poorest districts in Oklahoma:

Moffett

Crooked Oak

Talihina

Dahlonegah

Crutcho

Richest districts in Oklahoma:

Oakdale

Deer Creek

Merritt

Gove

Bixby

To the previous poster, Jenks is actually 8th richest, and Grant is only 141th-poorest. The problem you're going to have with this project is coming up with some way of assigning a 'running results' variable to each school district. If you only look at the finals for the state track meet, most school districts do not have anyone who made the finals, which is going to give you bad results in your regression.

I'd look in just your conference. Get data on average family income in each town. Then just look at the conference results and calculate points in the conference meet/town population (to control for bigger towns having a larger talent pool). Plot and run a regression on that shit in R or JMP or your calculator and find the correlation.

It would be interesting to focus your study on the state of Connecticut...a small state well known for income disparity. There are a small number of school districts, almost all organized by individual towns. It is very easy to find income information for these towns. For instance, see here:

http://ctboom.com/mind-blowing-gap-connecticuts-richest-poorest-school-districts/

)

I would also make sure that you control, or at least look at, school size as a key variable.

All Connecticut state meet well organized and easy to find at:

http://mysportsresults.com/Results/default.aspx

Data goes back at least 10 years.

My hypothesis is that there will be a relationship between wealth and cross country state finish. The relationship may exist in track, but it will be weaker.

Why don't you do something that's a little bit easier to collect data? Here's an idea. Pick a race you can get race results for. Maybe one you ran in. Then fit a regression (or ANCOVA) model that includes age and gender. You're probably going to want to include a quadratic term for age as well. Then using that equation predict where you were expected to run if you were joe average hobby-jogger (of your age and gender) and compare that to what you actually ran.

As part of this project you should summarize the race. What proportion were male/female. Give a distribution of the ages/age groups. Summarize the times by gender and age group. What was the average versus median time. And so forth.

By the way, you can easily convert the times in excel into minutes by using something like =A1*60*24. This assumes the cell in A1 contains the time as like 16:42. Excel will autoconvert this time into 16.7 but multiplying it by 60 mins and 24 hours which will give you the value you want for running the analysis.

Followup idea: find a race that's run on the same course in winter versus summer and compare results (race results in my town often list race day temperature so you could include that or mention it). Could even be like an XC meet early in season versus late in season run on same course.

One more thing. Whatever you decide to do make sure you discuss that times are often skewed for race data. Address the issue of potential skewness and outliers and how these might affect your results. Explore any outliers in your data and discuss what it means. Was it likely a data entry problem or just a bad result. I think this shows a more in depth understanding of what you're doing. As a statistician I hate when we just throw things into regression and forget about assumptions. I admit I do that myself sometimes and of course it can lead to biased results.

Poor districts may do better in sprints

Rich districts likely better at distance