Monmouth says Hillary has a 13% lead. This poll is major league amateur hour<\/p>\n
The poll<\/a><\/p>\n The 803 people sampled were pulled from two different datasets. The criteria is clearly they wanted 1\/2 landlines and 1\/2 Cell Phones, rather than the actual number.<\/p>\n With 803 sample of 142 million people, small judgement errors get magnified enormously. The survey “adjusted” for non representative sampling in party identification, sex, age and race. The weighting doesn’t adjust for things like section of the country, rural vs urban, unemployed (more likely to be reached by phone), number of children, marital status, income… The poll weighted on those attributes they considered significant.<\/p>\n The poll assumes 47% of the population is men. They actually got 51% men. The survey assumes that 25% of registered voters should be aged 18-34. What they actually got was 15%. Weighting is to some degree necessary, but the larger the deviation between the population and the sample, the less reliable the sample is. With 803 people, that means only 59 were women under 35. Their model predicted 106 women should have been under 35, so they needed to weigh those women almost 2:1… Since there are 4 weighting factors, all 4 have to be combined – not by addition or multiplication. In the under 35 women, what percentage are Republicans and how many did they get? Does the weighting apply equally on every question? If a black woman under 35 woman is weighted 2.5:1, and you ask her if she thinks president Obama is doing a good job as president, we already know that 86% of her sample will say yes, so 200% of all under 35 black women support Obama. See the problem?<\/p>\n The solution is to fix the sample, not to weight it after the fact.<\/p>\n","protected":false},"excerpt":{"rendered":" Monmouth says Hillary has a 13% lead. This poll is major league amateur hour The poll The 803 people sampled were pulled from two different datasets. The criteria is clearly they wanted 1\/2 landlines and 1\/2 Cell Phones, rather than … Continue reading