Test of significance for a series of statistical tests.

Behavioral and emotional problems among Chinese and American children: Parent and teacher reports for ages 6 to 13

“…Effects accounting for 1 to 5.9% of the variance are considered small, 5.9 to 13.8% medium, and > 13.8% large. Because some effects may be significant by chance, we have indicated with superscripts the smallest significant effects in each set of comparisons up to the proportion expected by chance, using a p < .05 protection level (Sakoda, Cohen, & Beall, 1954). The smallest effects are the ones that are most likely to be significant by chance.…”

Psychopathology in Dutch young adults: enduring or changeable?

“…After correction for chance findings (Sakoda et al 1954), using a P < 0.05 protection level, we found a higher average stability coefficient for the Thought Problems scale for females than for males (z = 3.19, P < 0.001). This difference reflected a difference for the age group 18-19years (z = 2.89, P < 0.005).…”

“…Applying a P < 0.01 protection level (Sakoda et al 1954), we controlled for chance findings by correcting for the number of comparisons. Effect sizes were judged according to Cohen's criteria (1988).…”

Consistent Differences in Individual Reactions to Drugs and Dummies

“…The number of tests actually observed to reach each level was 12, 6 and 2 respectively (Table IV: all the results attaining the 10% level or better are tabulated with others of interest). The likelihood of obtaining a constellation of scores such as this by chance is less than one in a thousand (Sakoda et al, 1954).…”

“…The mean number of symptoms/occasion reported for the dummy was indeed smaller than that to any drug except A, but the overall differences were insignificant. The number reported after taking drug F was significantly greater than that to the rest combined (P<0.02), but as a significance of this order could itself well be attained in one of the set by chance (P>0.1) when 6 comparisons are possible (Sakoda, Cohen, and Beall, 1954), the consistency of individual reports about the drug effects was examined on the assumption that the substances did not differ greatly in their pharmacological effects (Table III). The probability that the second reaction could be predicted from the first in any pair of successive occasions more accurately than could be expected by chance was on two occasions highly significant, as was the greater accuracy of the predictions summed over all occasions.…”