A Kaggle-esque competition for members of the BTV Data Science meetup group. An opportunity to develop your analytical skills, and earn some bragging rights.

The data consists of 250,000 records, each representing a traumatically injured patient who survived long enough to reach a hospital. Each observation has 1,000 binary predictors prefixed ais representing individual possible injuries (0=injury not present, 1=injury present) and a single binary outcome died, that is, patients who went on to die before discharge from the hospital (0=survived, 1=died).

The training and test data are available as a tar.gz archive here. Once you extract the archive, you'll have a statsathon directory containing the training (AIS_train70.csv) and test (AIS_test30.csv) data.

Details of the event are available on the Meetup page.

Feel free to participate by yourself, or form a team.

Analyze the data using whatever methods you want. To qualify for the Grand Prize of a free BTV Data Science sticker, you must use open-source software (e.g. R, Python) and make your code available, preferably in this GitHub repository.

Entries must be made by 9pm on Tuesday March 22nd. The preferred format for submission is by submitting a Pull Request to this respository (details below), with your predictions as a comma-delimited (.csv) file placed the model_evaluation directory, given your team's name. The first column of the csv file should be the ais_data_test30.INC_KEY in the test data, and the second column is your prediction of died, like this:

ais_data_test30.INC_KEY,pred
13000000,0
13000005,0
13000013,1
13000026,1
13000045,0
....

A complete example file, example1.csv, is in the model_evaluation directory. Alternately, you can email your results to [[email protected]].

All entries will be be scored against the test set, and submissions ranked by their accuracy, sensitivity, and specificity.

Please follow this Guide to using Pull Requests to submit your results and contribute your analysis. We're relying on the honor system that you don't look at any code that is submitted early :)

General outline:

1. Fork this repository to your personal account
2. Create a new branch with your group name (e.g. BTVawesome)
3. Create a new directory within the project using your group name (e.g. BTVAwesome_analysis)
4. Analyze data!
5. Include your scripts for analysis in your directory and commit code to your branch
6. Place final predictions as a csv (described above) with your team name (e.g. BTVawesome_pred.csv) in the model_evaluation directory
7. Submit a Pull Request to the BTVDataScience/statsathon-2016 repository