project

• Your preliminary proposal is due in Dropbox on Monday Nov 26 (preferably as a plain text file or R markdown); it should specify the question you want to ask, and ideally explain where you’re going to get your data from. The final version is due on Avenue on Monday December 10.
• the project should be R code with comments (.R), R markdown (.Rmd), or Sweave (.Rnw)
• none of the following functions should appear in your solution:
  • setwd()
  • rm(list=ls())
  • attach()
  • install.packages(...)
• the TA or I should be able run your code from scratch without any problems.

Goals

I want you to find a data set and answer some questions with it. Your writeup should consist of:

• a statement of the question(s) you’re trying to answer;
• a preliminary statement of the methods you initially planned to use (i.e., before snooping);
• some graphical explorations of the data;
• one or more model fits;
• graphical and quantitative diagnostics
• inference based on the model

• conclusions

• Please provide complete references specifying where your data came from, and any other sources you used.
• It will be best to pick a “medium-sized” problem/data set; if your example is too small (e.g. only one predictor variable, or a small number of observations that only for very simple models), you won’t be able to draw very many interesting conclusions.
• If you use a data set that’s taken from a fully worked example (e.g. this logistic regression example from the UCLA idre web site, you need to (1) reference the example and (2) explain clearly what you’ve done to differentiate your project from the worked example (did you use a different subset of predictors, consider different models …?)

• Please feel free to contact me by e-mail to discuss possible project ideas. If you have another idea (e.g. a simulation study of some modeling approach), let me know by e-mail.

Data sources

• Open Data Stack Exchange (for example, here’s a question on automobile insurance claims data)
• UCI Machine Learning Repository
• How to find and reuse data on Figshare
• Data Dryad (mostly biology)
• Scientific Data Repositories
• Open ICPSR (social science data)
• Quant Stack Exchange question on financial and economic data sources
• many scientific journals, such as PLOS ONE, require data underlying any paper they publish to be available and accessible
• if you find a graph that clearly displays the raw data you need, but the data are not presented in a numerical form, there are a variety of graphics-scraping tools available