Many are the abuses of statistics. That one you see almost every day is about polling. Sample size, polling, bias, conclusions are all in the mix.
A. The natural desire to use
correlation as causation. (Ubiquitous)
B. Unwitting bias in sampling and
survey structure. (Consider polls.)
C. Applying multiple statistical
tests until some significance is confirmed.
D. Using statistical tests with
sample sizes too small. (Chi-square test for example).
E. Using factor analysis while not
understanding what it does. (Looks good in a paper.)
F. Doing the study first and then
creating the null hypotheses*. (This is statistics backward. Is it ok if
everybody does it?)
G. Worst of all is taking
statistical significance as a form of proof. (Remember that at the 0.05 level,
5% of all significance conclusions are wrong. Combine with the thousands of
studies every year.)
*Personally guilty.
Comments
Post a Comment
Please Comment.