2. Use the word "heteroscedasticity" in conversation with a non-statistician. Extra credit if it triggers a serious long-term relationship.
1. Try the randomization in twoflowers.R using a different statistic. Feel free to be creative. If you want to use the same exact data from class, figure out how to use "read.table" to read in the data saved in pdat.txt. (If you want to experience a genuine, although optional, thrill, use 2*Sdiff (the difference in the log likelihoods) as your statistic and see if the distribution of randomized values follows a chi-squared distribution with 1 degree of freedom like it is supposed to).
2. Generate a new set of data where there are now three types of plants, two kinds of sagebrush and "other". Come up with likelihood ratio tests to compare models where all three plants differ, where all three are the same, or where the two sagebrush are the same but the "other" is different.
3. Write a short summary and emotive response to each of the seven tests we conducted on the data. Extra credit for haiku.
Find the program "quitscience.R" under rare disease stuff in the Rcode folder on the course web site. Experiment with the effect size and the significance level of the test.
1. Can you find values where the results look pretty good?
2. What's the worst you can do?
3. Use the "justsignif" column and come up with some way to compare the barely significant with the very significant results. Does it match what is claimed in the paper?
4. Figure out how to make some sort of graph that illustrates something interesting.