Figures don't Lie but Liars Figure
I'm relying on my memory for this, so I may be a little off on the exact numbers but the techniques I know cold having been asked to do "survey adjustments" several time in a past life.
Now to the point. A recent poll shows Hillary Clinton ahead of Donald Trump by 9 points. Normally I would be somewhat concerned by this, being a Trump supporter, but this is not about that this is about the survey and the underpinnings of the sample used. A little statistical theory : the results of a survey and its validity and veracity have much to do with the sample used. The number of individuals in the sample is almost but not quit irrelevant. It is the population stats that ater, not the size. If you choose 100% democrats and ask if they are going to vote to the Republican, then say that the Dem is ahead 70 to 30, your number are correct mathematically, but not valid. your sample was biased. Likewise if you go for the republican, mathematically correct but invalid.
To prevent this type of error, the survey must accurately reflect the population. In our case, the recent pool The Survey purports to be an accurate sample of the population, X % Democrat, Y% republican, and so forth. From the numbers I was able to find about the sample this survey was based on, there was a subtle error. The sample was based on the 2012 voter universe, not the 2016 voter universe. So what is the question. There is an over sampling of about 10% favoring the democrats. This means there is 10% more democrats that there should be. There a 6 % under sampling of republicans and a 2% under sampling of independents (all estimates approximately right).
If one removes the over sampling and adjusts the Republic and Independent percentages and then refigures the percentages, Mr. Trump is still behind, but only 2%, not 9 and is with the margin of error.
Knowing how meticulous survey firms are, I would hesitate to say the Democrat over sampling was deliberate, but it does tell me the polls are suspect since potentially all have over/under sampling issues and one needs to know what they are reading and attempting to make sense of so
POLL READER BEWARE!! and BE AWARE
Now to the point. A recent poll shows Hillary Clinton ahead of Donald Trump by 9 points. Normally I would be somewhat concerned by this, being a Trump supporter, but this is not about that this is about the survey and the underpinnings of the sample used. A little statistical theory : the results of a survey and its validity and veracity have much to do with the sample used. The number of individuals in the sample is almost but not quit irrelevant. It is the population stats that ater, not the size. If you choose 100% democrats and ask if they are going to vote to the Republican, then say that the Dem is ahead 70 to 30, your number are correct mathematically, but not valid. your sample was biased. Likewise if you go for the republican, mathematically correct but invalid.
To prevent this type of error, the survey must accurately reflect the population. In our case, the recent pool The Survey purports to be an accurate sample of the population, X % Democrat, Y% republican, and so forth. From the numbers I was able to find about the sample this survey was based on, there was a subtle error. The sample was based on the 2012 voter universe, not the 2016 voter universe. So what is the question. There is an over sampling of about 10% favoring the democrats. This means there is 10% more democrats that there should be. There a 6 % under sampling of republicans and a 2% under sampling of independents (all estimates approximately right).
If one removes the over sampling and adjusts the Republic and Independent percentages and then refigures the percentages, Mr. Trump is still behind, but only 2%, not 9 and is with the margin of error.
Knowing how meticulous survey firms are, I would hesitate to say the Democrat over sampling was deliberate, but it does tell me the polls are suspect since potentially all have over/under sampling issues and one needs to know what they are reading and attempting to make sense of so
POLL READER BEWARE!! and BE AWARE
