FAQ/addcounts62013-03-08 10:17:38localhostconverted to 1.6 markup52012-07-09 13:28:39PeterWatsonRevert to revision 1.42012-07-07 19:39:28bba428330.alshamil.net.aejltGxe , [url=http://udmrxlzmzhwb.com/]udmrxlzmzhwb[/url], [link=http://yfzmyepwvbnm.com/]yfzmyepwvbnm[/link], http://pshzvzslpkoy.com/32012-07-07 14:03:11sa-84-52-29-136.saturn.infonet.eevrVm0M <a href="http://sypufouoosmw.com/">sypufouoosmw</a>22012-07-06 21:33:56188.143.232.12The 1/sqrt(n) approximation of two strndaad deviations deteriorates pretty quickly as p moves away from 0.5. It's off by nearly a factor of 2 for p = 0.9 as indicated in your table. [Also, you nee12008-11-11 16:21:27PeterWatsonHow do I combine proportions?Proportions which measure the same underlying concept, such as attention, over blocks of results can be pooled by simply obtaining an aggregate sum of the hit rates. Pooled estimate = $$\frac{\mbox{total number of correct responses}}{\mbox{Total number of possible correct responses}}$$ So for example if 5 out of the 16 questions were correctly answered in the first session and 17 out of 25 in the second session the pooled estimate would be (5+17)/(16+25)= 22/41. This pooled proportion is a sum of observed frequencies. Note Do not sum percentages to obtain a pooled estimate because a percentage does not take into account the sample sizes (total possible number correct). e.g. 1 correct out of 2 and 500 out of 1000 both have 50% correct but we would have more confidence in the latter as it is based on a far larger sample size. This is why for analysis purposes, such as chi-square tests, the observed frequencies are used rather than percentages.