Picking the Right Statistical Test
Written by Robert Niles
Congratulations! Most journalists have no clue that there are different tests for different situations.
Here's the best advice I can give you: Talk to a pro. You know how great copy editors can catch errors in syntax, usage and vocabulary that even experienced writers rarely notice? Professional statisticians do the same thing with numbers.
Smart reporters run their words by a copy editor before they hit print. Why not run your data past a statistician before publishing them?
Unfortunately, I don't know of many publications that have people with statistics degrees on their editorial staff. While managers might not feel that correct numbers and proper analysis are important, our readers do. And the screw-ups that our collective lack of attention to things like statistics has caused might be part of the reason why readership's dropping at so many U.S. newspapers.
Call the press relations department of your local college or university and ask for a contact in the statistics department. Then talk with that source about what you have and what you want to do. As with any source, it's best to establish the relationship off deadline, when you have time to ask questions and wait for thoughtful answers.
"Okay, that's nice, Robert," you say. "Um, I'm under deadline now for this story/article/paper/homework assignment, and really need to know what to do...." Well, then, here are some tips:
The best resource I've found for figuring out the right test to run is Selecting Statistics, from Bill Trochim at Cornell University. To use this site, you'll need to know a little bit about your data. The site will ask you a series of questions about your data, and pick the right test for you, based on your answers.
If you want to understand why a specific test is the right choice, try Intuitive Biostatistics: Choosing a statistical test, an online chapter to a stats textbook.
When you're ready to conduct your test, you'll find links to several nifty web pages that perform stats calculations at statpages.org. And additional information on testing can be found at David Lane's HyperStat Online.Finally, if you still need more help, may I suggest one of the books on my recommended reading list?