"-rick-" wrote in message
. ..

"NOYB" wrote in message
ink.net...

Yeah, they conveniently leave out the statistics about how many
respondents were registered Democrats vs. registered Republicans. But
the AP-Ipsos poll *did* provide that data: 52% registered Dems and 39%
registered Republicans. I'd suspect that this poll is just as badly
skewed.


You misstate the data.

----
Question #1:

Do your beliefs tend to lean more toward the Democrats or the Republicans?
----

Which does not necessarily equate to party affiliation.

" Half the respondents were asked party identification in this location, the
other half were asked at the end of the survey with the

other demographic questions. Results for the respondents who were asked the
question early were 40% Republican, 50%

Democrat. For the respondents asked later in the survey, the results were
39% Republican, 52% Democrat."





(In your opinion, party identification doesn't mean the same thing as party
affiliation? )




(hint: independents)

Yes, independents...which is explained by the fact that 52%+39% doesn't
equal 100%.

"NOYB" wrote ...

" Half the respondents were asked party identification in this location, the
other half were asked at the end of the survey with the

other demographic questions. Results for the respondents who were asked the
question early were 40% Republican, 50%

Democrat. For the respondents asked later in the survey, the results were 39%
Republican, 52% Democrat."

(In your opinion, party identification doesn't mean the same thing as party
affiliation? )

Not exactly. Expectations and performance vary over time. "Do your beliefs
tend to lean more toward the Democrats or the Republicans?" is not equivalent to
"what is your party affiliation" or "are you a registered republican or
democrat?" as you seem to imply. That was my point.

I don't mean to unreasonably split hairs but your premise that this response
invalidates the poll isn't obvious to me. If a statistically significant random
sample yields that result why is it not valid within the probabilities of it's
margin of error? Do you have evidence of non-random sampling?

-rick-

"-rick-" wrote in message
...

"NOYB" wrote ...

" Half the respondents were asked party identification in this location,
the other half were asked at the end of the survey with the

other demographic questions. Results for the respondents who were asked
the question early were 40% Republican, 50%

Democrat. For the respondents asked later in the survey, the results were
39% Republican, 52% Democrat."

(In your opinion, party identification doesn't mean the same thing as
party affiliation? )

Not exactly. Expectations and performance vary over time. "Do your
beliefs tend to lean more toward the Democrats or the Republicans?" is not
equivalent to "what is your party affiliation" or "are you a registered
republican or democrat?" as you seem to imply. That was my point.

I don't mean to unreasonably split hairs but your premise that this
response invalidates the poll isn't obvious to me. If a statistically
significant random sample yields that result why is it not valid within
the probabilities of it's margin of error? Do you have evidence of
non-random sampling?

The sampling may be fine. It's the weighting that I have a problem with.