re: #124 JamesWI
Well, the unanimous thing was just a simple example. But 57 seems like it could be considered too few to have an accurate breakdown. The margin of error on a group that small would be huge.
It’s big but publishable. If your population for Hispanic voters is 9.7 million as it was in 2008 then the margin of error on a sample size of 57 is 13. It also raises the point that such under sampling is itself a bad practice, especially when the percentage being under represented is statistically significant.
Corrected above 2008 Hispanic vote was ~7.4% of total votes cast. Does anybody really think that percentage is going down this year? Look at that percentage by the old breakdown 77% obama = 5.56%. Now take out Romney’s portion and there’s a 1.7% shift implied in omitting Hispanics from the final tally.