Due - 4:30 PM Tuesday, November 24th.
You should have chosen a data set and procedure by Tuesday, October 27th. You must get your data set and the procedure you wish to use approved in advance. You must also get the plan for your simulation study approved in advance.
The purpose of this project will be to explore how the nonparametric statistical procedures discussed in the course can be applied to either the data in your field of study, or data that you are interested in. You do NOT need to collect the data yourself. I will provide assistance and guidance in any issues involving the use of Splus to complete the project.
If you pick a particularly complicated data set it may be necessary to only deal with a subset of that data set so that we can use the procedures that we have covered. HOWEVER, I will be happy to meet with anyone who wishes to use a nonparametric technique that we have not yet reached. A list of procedures and their usage can be found from pages 10 to 14 in the text.
The final report must be typed and should form a cohesive whole, with transitions between each of the five parts listed below.
Except for part IV, the paper should be addressed to an audience whose members may not have much statistical training.
Make sure that every issue raised below for each part is addressed.
Points may be deducted for grammatical and spelling mistakes.
Part I - Give a summary of the data set including any needed plots or graphs. This summary should include: background on why the data set was collected, why the data set is of interest, what hypothesis is desired to be checked, and the basic summary statistics and plots for the data set. The justification of why the data set is of interest should include supporting information from at least one source in addition to the one containing the data set. Cite all references appropriately.
Part II - Provide a brief overview of the assumptions required for the nonparametric test of the hypothesis in Part I, and also for its parametric counterpart. Conduct and report the results of the tests and plots needed to verify these hypotheses. This overview should be understandable to someone without much statistical training. Summarize which of the nonparametric or parametric tests is valid based on these results.
Part III - Carry out BOTH the nonparametric test and parametric test on the data set, being careful to note which (if any) of the tests are invalid because their assumptions are not met. Give the computer code you used for the analysis in an appendix to the report. Report the p-values, measures of effect size, and confidence interval for the measures of effect size. Summarize these statistical results in a manner that is understandable to someone who does not have much statistical training.
Part IV - Conduct a simulation study of the power and nominal type one error rates for the parametric and nonparametric procedures you have chosen on the same sample size as the data set you have chosen. The distributions used for the power study will depend on your particular data set, and must be approved by me. Provide the code you used for your simulation study in an appendix. In the main part of this section of the report, begin with an introduction stating that the section is very technical and can be skipped by those not statistically inclined. Summarize what you are attempting to accomplish with the simulation study. Report how well the procedures keep to the nominal type I error, and how their powers compare.
Part V - Provide a summary tying together the results of the previous four sections, and what use (if any) nonparametric statistical techniques have for those analyzing data sets similar to the one you analyzed.