A real improvement or a random noise?

[Shira]

If you look at anything in Rasmussen, look at the 7 day average.

OK. I took the numbers from your table.
During the period 9/12-9/18 the avarage was 48.5:45.3 (a 3.2% difference) in Bush's favor.
The numbers in the next 7 days between  9/19 and 9/25 are 48:45.8 which is 2.2% difference.
This means that Kerry REALLY improved by 1% in the last week. These are seven days samples with very small statistical noise.
The methodology of Rasmussen could be bad but it is the same methodology in the compared two periods.

[The Vorlon]

a 1% change between 2 samples with N=7000 has a 61% chance of representing an actual change - a 39% chance of just being "noise"

[CARLHAYDEN]

Are these day to day numbers all already weighted 39-35- 26?

yes, the single day vaues are weighted to 39/35/26.  I can show you the unweighted values as well, but the bounce around a ton and are not particularly useful.

Rasmussen actually has a rather large "Day of the week" issue.  He weights a ton of it away with his 39/35/26, but in the raw data it is rather a meaningful effect.

As you know, I do believe in partisan weighting (but also believes it changes over time).

Right now I see the following:

Democrats          38%

Republicans         37

Others                25

Would appreciate your feedback.

[Smash255]

Are these day to day numbers all already weighted 39-35- 26?

I still see it about the same as last election wtih the 4% Dem ID advantage.  A 1 % or 2 difference from that is possible, but I don't see it any larger than that

yes, the single day vaues are weighted to 39/35/26.  I can show you the unweighted values as well, but the bounce around a ton and are not particularly useful.

Rasmussen actually has a rather large "Day of the week" issue.  He weights a ton of it away with his 39/35/26, but in the raw data it is rather a meaningful effect.

As you know, I do believe in partisan weighting (but also believes it changes over time).

Right now I see the following:

Democrats          38%

Republicans         37

Others                25

Would appreciate your feedback.

[The Vorlon]

Are these day to day numbers all already weighted 39-35- 26?

yes, the single day vaues are weighted to 39/35/26.  I can show you the unweighted values as well, but the bounce around a ton and are not particularly useful.

Rasmussen actually has a rather large "Day of the week" issue.  He weights a ton of it away with his 39/35/26, but in the raw data it is rather a meaningful effect.

As you know, I do believe in partisan weighting (but also believes it changes over time).

Right now I see the following:

Democrats          38%

Republicans         37

Others                25

Would appreciate your feedback.

I am crunching data to write a nice post.. wait for it

[agcatter]

looking forward to it