Tuesday, 21 May 2013

"Don’t rely on statistics to prove that the effect is real" by Philip Teitelbaum


The following one, is an extract from Philip Teitelbaum' s paper:

Teitelbaum, P. (2012). Some useful insights for graduate students beginning their research in physiological psychology: anecdotes and attitudes. Behavioural brain research, 231(2), 234-49. doi:10.1016/j.bbr.2012.01.030

The paragraph briefly presents the author’s view on statistics in physiological psychology research.

Enjoy!



“Working on a big phenomenon often eliminates the need to use statistical analysis to decide whether the treatment has produced a reliable effect: this can be obvious in each individual subject. As Smitty Stevens, the great psychophysicist who worked on hearing at Harvard, used to say, “Don’t work down in the noise level.” When I was an instructor at Harvard, I served on Paul Rozin’s advisory committee. He wanted a joint Biology–Psychology degree, so his committee included some biologists. The first thing the psychologists suggested for Rozin to study was statistics. However, a biologist said, “If he learns statistics, he will never do anything good in science.” This echoes the thinking of Claude Bernard [1,2]. My own view on this matter is that if you need inferential statistics to decide whether you have made a difference with your treatment, your method is inadequate. At this point, I have always taken the approach that I must improve my method of measurement or pick a bigger phenomenon to work on. A big phenomenon thus facilitates rapid research. For each experiment, you need to replicate the result in only a few animals or people, rather than having to do so in many to determine statistically whether the variable is having a significant effect. This allows you to vary your method in experiment after experiment, which is far more important in attaining understanding than is the use of many subjects whose results are averaged to produce statistical significance. Personally, I do not depend on statistics to convince me that I have found a real phenomenon. One animal or person is all it should take to do so. Sometimes it is difficult to reproduce a phenomenon after I have seen it once. If it is big, however, I know it is real, and that it is important. That is one reason I try always to record each new puzzling, phenomenon on film or video at the time I see it. The video encourages me to keep trying to reproduce it for further analysis”.

References
[1] Bernard C. An introduction to the study of experimental medicine. New York: Dover; 1865 [reprinted in English translation, 1957].
[2] Bolles RC. Why you should avoid statistics. Biological Psychaitry 1988;23:79–85.


No comments:

Post a Comment