Tartz et al. (2007) performed two-sample z-tests for proportions comparing men and women on 28 different indices of dream content. They make the following comment:

When making 28 comparisons, an average of one comparison might have been significant at the .05 alpha level by chance alone (i.e., a false positive or type I error). Actual results yielded two significant results, one more was close to significance…

(a) Give a more precise calculation of the expected number of type I errors.

(b) How could the authors have controlled the chance of type I error so that the entire list of 28 comparisons had only a 0.05 chance of any false positives?

(c) The two results they cite as significant had p values of 0.004 and 0.007, respectively.Would these have still been significant if we adopt the strategy you suggest? LO.1