Using Confidence Intervals

This website is dedicated to gathering information on the understanding and use of confidence intervals in reporting experimental data. It was finally inspired by a seminar led by Michael Smithson on 15 Nov. 2000. I am relatively inexpert, so I shall mostly refer people to other pages. When I find a website that fulfills my goals better than my own page, I shall turn this into an automatic forwarding site.

I discover that I have not updated this page in over a year, and this is unlikely to change soon unless people ask. So either bug me or show me a better page so I can just forward people there.

General

Right now the most useful page I have seen belongs to Bruce Thompson and can be found at: http://acs.tamu.edu/~bbt6147/. It concentrates on Effect Size reporting, and includes plenty of references and links to online sources.

Reading on CIs

Two articles suggested by Smithson for an immediate introduction to confidence intervals.

• For a "nontechnical, if passionate, article on confidence intervals by one of their best exponents":
  Frank Schmidt. 1996. Statistical significance testing and cumulative knowledge in psychology: Implications for the training of researchers.Psychological methods 1:2 115-129.(based on his Presidential Address in APA division 5)

• For an article on noncentral CI's, see:
  Steiger, J.H. and Fouladi, R.T. 1997. Noncentrality interval estimation and the evaluation of statistical models. In L. Harlow, S. Mulaik, and J. Steiger (Eds.) What if there were no significance tests? New Jersey: Lawrence Erlbaum

And of course, Smithson's new textbook:

• Michael Smithson. 2000. Statistics with confidence

Other Websites

Here is a list of useful sites I have found in my meager search. I've left out pages which obviously confused significance and effect size, or recommended against computing/using CIs if the effect was not statistically significant.

• A paper explaining CI's to clinicians: http://www.cma.ca/cmaj/vol-152/0169.htm It includes a discussion of estimation using a simple coin toss, and gives a feel for how wide a 95% bar will be.
• Are confidence intervals useful? Looking beyond the simple cases http://www.educ.ubc.ca/faculty/zumbo/ci/ci.htm Especially useful is the author's Table 1 which compares 3 (pre-1997) resources for computing and using CIs. (The author's links are broken, but you may get all the images by putting the linked filename onto the end of the main document's URL: e.g. ".../ci/filename.jpg"
• Confidence Intervals: http://stat.tamu.edu/stat30x/notes/node101.html A list of 11 kinds of CIs, after a brief introduction.
• WorldMedicus lists several articles on the use of CIs. There are at least 5 papers published in 2000! WorldMedicus on CIs

How to get papers using CIs published

• Read a Letter from Bruce Thompson on the subject.
• Read the APA taskforce recommendations

Demonstrations and Tools

• Geoff Cumming has an amazing set of Excel spreadsheets, which I hope to link to when they are more generally available.

Examples of papers using CIs well

Yes, there are some out there. I just haven't dug them out myself. Please feel free to submit any references. I'll even see if I can find some available online (e.g. jstor)

Examples of mistakes with CIs

The goal is to have a set of "don't-do-this" examples. Things to look for:

• failing to disclose what kind of CI was used
• failing to refer to or use the CIs that were calculated
• using the CIs only to perform null-hypothesis significance testing
• computing the wrong CI