it seems that all the existing prop test assume 2 independent or matched sample.but in the real world, many situations are not as we assume.for example,i do a research on the voter's prefernce through a random sampling.and the sample shows that 23% of the sample choose A,28% choose B,the others choose C.and i want to test the diference between the proportion choosing A and B (23%-28%=-5%) is sinificant or just due to sampling error. i think the prop.test is not proper here. and it seems all the stat textbooks do not deal with these problem. is it a igorance in the statistical world or just something else?
a statistic question,a bit off-topic,but important
2 messages · ronggui, Achim Zeileis
On Wed, 20 Apr 2005 01:36:02 +0800 ronggui wrote:
it seems that all the existing prop test assume 2 independent or matched sample.but in the real world, many situations are not as we assume.for example,i do a research on the voter's prefernce through a random sampling.and the sample shows that 23% of the sample choose A,28% choose B,the others choose C.and i want to test the diference between the proportion choosing A and B (23%-28%=-5%) is sinificant or just due to sampling error. i think the prop.test is not proper here. and it seems all the stat textbooks do not deal with these problem. is it a igorance in the statistical world or just something else?
If I understand you correctly, you're not interested in C at all and just want to test P(A | C') = P(B | C') = 0.5. This can be done by binom.test. Assuming the above were not proportions but observations binom.test(23, 23+28) Instead of performing this exact test you can also perform the asymptotic test via prop.test() prop.test(23, 23+28) Z
______________________________________________ R-help at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html