Hi Michael,

It is not a proportion, but an odds, which can vary from 0 to infinite. Even so the binomial test is apropriate? Another point is that at first the males are not independent of the entire population because they are part or the entire population. This isn't a problem?

My questions are due to the fact that I am also interested in this issue. I am used to constructing confidence interval for the odds ratio between males and females, if the interval does not include the value one, then the odds ratio is significant and this is usually everything I need.

Marcos

-----Mensagem original----- De: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] Em nome de Michael Kruger Enviada em: sexta-feira, 23 de abril de 2004 11:06 Para: SPSSX-L@LISTSERV.UGA.EDU Assunto: Re: stats question on odds

Paul Mcgeoghan wrote:

>Hi, > >I have a stats question which I wonder someone can help with. > >A paper states that the ratio of men who use alcohol hazardously >(1:1.2) was considerably higher than in the population (1:2.5). > >I have been asked to perform an appropriate test to explain whether or >not this is a statistically valid conclusion. > >In the sample, there were 200 people in total (53% of men used alcohol >hazardously and 44% of women). > >What do I use? > >Thanks for any replies, >Paul > > > Paul, The simplest test of your hypothesis as to whether this difference is statistically signifcant would be the binomial test. You have a proportion of 0.46 for males (1/2.2) and a proportion of 0.29 for the entirepopulation (1/3.5). The binomial test would test the hypothesis that the proportion for males is significantly different thatn the proportion for the overall poplation. The information you give with respect to sample size isn't nparticularily useful since we don't know how many males and females are sampled although you could determine the number based on the percentages given.

-- Michael Kruger "A True Prince" Statistical Analyst C.S. Mott Center Dept. of OB/GYN Wayne State University School of Medicine (313)-577-1794