Links for Days 3 and 4
Day 3:
Resampling Statistics software website. http://www.lock5stat.com/
The software is StatKey. At the main URL, find the datasets
for the text (you need the csv file to upload to the StatKey software),
some help documents, and a link to StatKey.
page 12. Cost-Benefit analysis for medical studies.
o
http://andrewgelman.com/2015/09/17/26612/
o
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2641547
o
http://content.healthaffairs.org/content/24/1/93.full
Power
calculations: https://istats.shinyapps.io/power/
[use population mean]for an interactive
demonstration. [Try mean 1, st deviations 1 and 5]
The File Drawer Problem Cartoon https://mchankins.files.wordpress.com/2013/04/filedrawer1.jpg?w=940
Martha Smith's blog http://www.ma.utexas.edu/blogs/mks
Day 4:
Multiple Inferences:
- Jerry Dallal's
Simulation of Multiple Testing. Suppose all the null hypotheses
are true. And suppose we test correctly using a 5% significance
level. What results do we get? http://www.jerrydallal.com/LHSP/multtest.htm
This simulates the results of 100
independent hypothesis tests, each at 0.05 significance level. Click
the "test/clear" button to see the results of one set of 100
tests (that is, for one set of tests all done in the same lab, or the
same year, or some other such collection of them.). Click the button
two more
times (first to clear and then to do another simulation) to see the
results of another set of 100 tests (i.e., for another researcher, or
another year, etc.).
Notice as you continue to do this that i) usually some of the tests
give "significant" results, even though, in fact, all Ho's are true, so
these are all giving you Type I error. Jerry Dallal's funny
examples : http://www.jerrydallal.com/LHSP/jellybean.htm
, http://www.jerrydallal.com/LHSP/cellphone.htm
- Jelly Beans (a cartoon): https://xkcd.com/882/
Data Snooping:
- Tyler Vigen's website: http://tylervigen.com/
or, for a longer list
http://tylervigen.com/old-version.html . (I don't think
that the older version described on the slide is still visible, but the
second website here allows you to take one variable and then choose
from a drop-down list some other variable to use to see what the
correlation is.)
Suggested from the class: