Zogby's single night numbers?

[lsuguy]

Could anyone pass along the individual night numbers for the last 3 or 4 days from Zogby? I'm constructing a spreadsheet that will estimate the single-night numbers from four different tracking polls (based on their 3-day rolling averages) so I can post that average here.

Thanks!

[Goldmund]

1. You can figure it out

It's a nice little math problem.

[papau]

2. Some of us are too old and use MathCad - but I grant you it is not

too tough to get to a "close" answer.

[Goldmund]

3. Do you have a link to the last 10 or so days of Zogby tracking?

I'll try to figure it out for you.

[papau]

4. here is the data

.

Reuters/Zogby Tracking Poll. Rolling 3-day sample. N=approx. 1,200 likely voters nationwide. MoE ± 2.9. Sponsored by Reuters, as of Oct. 4-6.


General Election Trial Heat:

Bush Kerry Nader Badnarik Peroutka Cobb Other (vol.)/
Unsure
% % % % % % %
10/15-17/04 45 45 1 - - 1 8
10/14-16/04 46 44 1 - - 1 8
10/13-15/04 48 44 1 - - - 7
10/12-14/04 48 44 1 1 - - 7
10/11-13/04 46 45 1 1 - - 7
10/10-12/04 45 45 2 1 - - 8
10/9-11/04 45 45 2 - - - 8
10/8-10/04 44 47 2 - - - 7
10/7-9/04 45 46 1 - - - 8
10/6-8/04 45 46 1 - - - 7
10/5-7/04 46 45 2 - - - 7
10/4-6/04 46 44 2 - - - 8

10/1-3/04 46 43 2 1 - - 9

[Goldmund]

5. Cool

So, let's concentrate, on, say, Kerry's rating.

The poll of 10/15-17 gives him 45%. That's the average of what he got on Oct 15, 16 and 17 respectively. Let's mark these ratings R15, R16 and R17.

We know that (R15 + R16 + R17)/3 = 45.
Therefore,

R15 + R16 + R17 = 135

Similarly, (R14 + R15 + R16)/3 = 44. So,

R14 + R15 + R16 = 132

Now, let's look at (R15 + R16) as one variable - call it X.

X + R17 = 135, and X + R14 = 132.

So, now we also know that

R17 = R14 + 3

In other words, his rating on the 17th was higher than on the 14th by 3 points.

Let's take another sample:

(R13 + R14 + R15)/3 = 44.

Therefore,

R13 + R14 + R15 = 132

Analogously to what we did last time (take (R14 + R15) as a single variable), we can conclude that

R13 = R16

Repeat the process once more for the 10/12-14 sample, and you'll find out that

R12 = R15


....


You know what man? I was wrong. I've been fucking my brain with this for a few hours now. The problem is that whenever you add an equation to the system, you add a new variable; you can always tell the difference between a certain day and a day 3 days earlier or later, but you can't specifically figure out what exact data is associated with a day. If we had specific data for any day, and for any day not distant from that day by a number of days divisible by 3, we could figure out the whole system. Like, if we had Oct 11th and Oct 15th, we could figure out every day. But this way, we're shit out of luck. Of course, you can make educated guesses. Like, if we assume that Oct 14th and Oct 15th are both 44, then the 17th is 47, and the 13th is also 44, etc.

[Awsi Dooger]

7. Do a search for a thread by TruthIsAll

I would help but I'm not a donor so I can't search.

On Sunday night/wee hours Monday morning TruthIsAll posted a thread in this forum that listed Zogby's daily numbers in the middle of the chart. It was entitled something like, "Prediction: Zobgy will have Kerry 47 -44 tomorrow"

[Goldmund]

8. Can't find the thread...

But also, the prediction was obviously wrong. Zogby had 45-45 on Monday, as well as today.

[Teaser]

6. Well, with caveats...

You have to assume some starting values for them and if you're off on the starting values, that can throw off the entire calculation.

With enough computing power, you can put confidence intervals on reverse engineering the data, but I wouldn't want to be the one to do it.

[papau]

9. Which is what MathCad and other great Math programs do!

:-)