[FourtySeven]

Dear Rage Monkey:

I posted this on another thread with no response that I could see, I’m genuinely curious if this has been taken into consideration. Please excuse any wet towels....

I hate to throw a wet towel on any of this but something has always struck me when claims of oversampling are made to make a poll look good (for us):

Isn’t it reasonable to assume that when a particular group is oversampled, the pollster in question takes that into account, and gives the group undersampled more weight?

For example, if the Republicans are undersampled by say 5%, their final tally is given 5% more weight than the Democrats.

That’s how it’s done in certain areas of science I can attest. For example, if one particular protein shows more expression in a certain experiment on a Western blot, such expression levels are normalized against a more common protein like actin. If the actin levels demonstrate lower levels in the experimental conditions (an oversampling if you will of the experimental condition), then the experimental level is reported as a function of the actin, that is, it’s divided by the amount of actin (similarly for the control), so as to normalize all levels based on how much total protein was actually loaded onto the gel (the Western blot).

So again, isn’t it reasonable to assume these pollsters are doing a similar kind of normalization?

[jps098]

That is incorrect. All but a few poll samples do not 'weight' for party identification. Rasmussen is one that does. In theory, if they only call from a list of Democrat donors, they would consider it fine.

Their assumption would be that if no one admits to being a Republican, then the country has shifted left. The problem is that with only 9 percent even answering the phone, the pool of respondents is small and getting smaller.

Yesterday, a Freeper posted a listing of the Methodologies used by all the major poll companies, and only 3 of about 15 even care to ask Party ID as a qualifying question. It is the same class of question as the person's age, just a demographic tally.