Resampling Stats Wizards
A drug company has come out with a new acne medicine that it claims is 90% effective in eliminating pimples. You apply it to 10 pimples, and only 7 are eliminated. Is this consistent with the company's claim?
The resampling procedure:
- Put nine "1's" (cures) and one "0" (no cure) in a hat
- Shuffle the hat, draw a number at random, record its value, replace the number
- Repeat step 2 for a total of 10 draws (corresponds to the application to 10 pimples)
- Note the number of "1's"
- Repeat steps 2-4 (say) 1000 times
- Find out how often we got <=7 "1's"
Using Resampling Stats software:
Use commands to define one simulation experiment, corresponding to steps 1-4 above, as follows:
This translates as follows: Using the URN command, place 9 "1's" and one "0" in a vector (list of numbers) called "hat." SAMPLE 10 values from "hat" with replacement, call the 10 resampled values "samp." COUNT how many "1's" there were, place the count in "cures."
Resampling Stats Wizards then handle the task of repeated simulation and analysis of results:
Click "proceed" and a dialog box appears to analyze the results:
We see that 73 out of the 1000 simulations saw 7 or fewer cures, when testing a random device that had a 90% "cure" rate. Using the traditional significance level of .05, the observed result can happen just often enough by chance so that we can't reject the company's claim.
More Examples
