Another proportion question-help much appreciated

[naivesincerity]

Say I know 3000 people were canvassed by a political party, 2200 of them went out to vote. Of the 7000 not canvassed, 4000 went to vote.


What kind of test should I do to say something about the population proportions of people who vote according to whether or not they're canvassed?

Many thanks, ns

 

[BioStatMatt]

You could use a 2X2 contingency table here. Where canvassed/not-canvassed corresponds to the rows and voted/not-voted corresponds to the columns. You could then use a test for homogeneity, but I would prefer to calculate the odds ratio. An appropriate interpretation of the odds ratio would allow you to make a statement like the one you want. You could also calculate a confidence interval around the odds-ratio.

~Matt

 

[naivesincerity]

Thanks BSM