Exercise 2

Due date : 30/9 before 21:00, via email. Send a short pdf with your code, results and discussion.

Can be done individually or in pairs

Small data set 1

File one.csv contain two sets of paired data (each set is a colum, the pairing is the line).. Run the following 2 paired sample tests on then and report the p-value.

• t test
• Wilcoxon signed rank test
• a Monte Carlo permutation test (with 2000 random permutations/replicates). this discussion in the Q&A site Cross Validated has a discussion on permutation tests for paired data and the code in R.
• The sign test.

Does the order of p.values correspond to what one would expect given the power of the tests (assuming the data is gaussian like)?

Run the t-test and the Wilcoxon without using the paired information. Did it cause large changes of the p=value?

Binary data set 2

File two.csv contains a paired binary data (coded as 0/1). Run the following precedures and compare the p-values. Discuss if it is appropriate or not to run the test

• McNemar test
• (non paired) Chi-square test
• paired t test
• Wilcoxon signed rank test