comparing percentage when without knowing n? what test?

#1
I'm trying to do a meta analysis of some poll results and though I know the sample size in general, if I'm comparing the reported results and they say things like "78% of Toyota owners like baseball" but they don't say which percentage of people overall were Toyota owners, can I still test for significance of those percentages between surveys? Like if I want to look at a series of polls and compare how much Toyota owners like Baseball compares to Honda owners, can I do this and test for significance without knowing the actual number. If so, how? Thanks!!!!

whoops -- typo in the subject line and i'm now allowed to edit it. srry.
 
Last edited:
#2
Hi,
if you do not know the sample size, you cannot do any test with percentages. Because a single % does not give you any idea on the variability. And stats is about variability and potential errors you make when estimating a statistic when having a sample.
Camille
 

trinker

ggplot2orBust
#3
I usually do not comment with I agree statements without having something else to add so...
I agree with CamilleJosion especially when you consider Simpson's Paradox (Thanks for bringing that to my attention Dason). This paradox illustrates just how tricky percents can be.