Rasmussen daily graph for 12/14/07 - Clinton and Edwards up 1, Obama down 1

[DavidDvorkin]

(Pretend that the 10 and 20 lines are actually 0. Does anyone know how to make Excel label different parts of the axis differently?)

Rasmussen links:
Data in tabular form
Discussion

[denem]

1. The fix is in

obviously. Thanks again for your work. Greatly appreciated.

[TwilightZone]

2. How long a time frame is the extrapolation based on?

Hillary's recent semi-recovery seems to be underrepresented in it. The extrapolation shows her trailing off, but the recent data doesn't seem to support that.

[DavidDvorkin]

5. It's a one-week extrapolation

But the trend lines are a subjective choice on my part, in a sense. (I should probably add something about that to the post, so that it always shows up.)

I played around with different types of trend lines and settled on a second-order polynomial fit because it responds to recent data but doesn't overrespond. But that's all it is - it isn't revealing some kind of underlying process, the way fits to physical data may do.

If a real, long-term trend develops for one of the candidates, then the trend line will give us an idea of what will happen if that trend continues.

[TwilightZone]

7. Sorry, I wasn't clear in my question.

I knew that the extrapolation was one week, since it's labeled. My question really pertained to the trend. It looked like a moving average (that was extrapolated), so I was curious to see how long the "rolling" period was, if that was the case.

Long-term trends may not tell us much, considering the historic last-minute volatility in a lot of places - in Iowa, in particular. Someone else posted an article that said that something like 60% of Iowans didn't choose a candidate in '04 until the last week.

Thanks for the follow-up and for tracking this info in the first place. I always find it interesting.

[DavidDvorkin]

8. Oh, I see

It covers all the data. I've been thinking that it should start fairly recently, instead, and I may play with that to see what difference it makes.

[TwilightZone]

9. Thanks again.

I'm a math mind, so stuff like this always makes me a little overanalytical. :)

[DemocratSinceBirth]

3. Two Small Points

I am suspicious of Rasmussen's polls and a lot of these small movements are statistical noise...

This doesn't reflect on your work which is great...

[DavidDvorkin]

6. They're four-day rolling averages

Which smooths out a lot of noise that we don't even see. But I agree that a point or two up or down doesn't mean anything, as we can see from how the candidates then settle back to where they were.

[Lucinda]

4. Interesting stuff DavidD thanks! A couple of links for the excel info:

Its been a thousand years since I used excel and it's not on this computer so I can't tinker. Here are a couple links you might check out re:the axis question:

http://www.ozgrid.com/Excel/highlight-axis.htm
http://blogs.msdn.com/excel/search.aspx?q=multiple+labels+for+an+axis&p=1

[BootinUp]

10. suggestion

it might be time to lop off some of the early months, might give a better view of whats happening, maybe start from September. :)

[DavidDvorkin]

11. Yes, I'll probably do that

Maybe with tomorrow's posting.