OMFG! Please check my math.

[El Barto]

Rasmussen (http://www.rasmussenreports.com/Presidential_Tracking_Poll.htm) shows a sudden Bush surge today. This is the tracking poll that has been consistently at a 1 to 2% Bush edge for weeks. Today the edge is 4.6%.

But wait, this is a 3 day rolling average poll, so to produce such a bounce after a long stable period of results, last night's one-day results must have been a huge Bush surge!

My calculations:

Bush (47.2 + 47.2 + 53.5)//3 = 49.3

Kerry (46.4 + 46.4 + 41.3)/3 = 44.7

The 47.2 and 46.4 numbers are the average results for the previous 3 days which have only varied by a tiny .1 during that period. The 49.4 and 44.7 numbers are today's 3 day rolling average.

This suggests that last night Bush got a 53 to 41 edge in this poll that has consistently shown a 1 to 2 point race. (The one-day results only, before being averaged into the 3-day rolling average)

These are rough non-exact numbers, but my math is correct, yes?

[tanda]

Assuming that the result that dropped off was 47.2/46.4.

If the previous poll was
43.2 - 49.2 - 49.2 for Bush and
50.4 - 44.4 - 44.4 for Kerry
then a 49.5/45.5 yesterday would get the current results.

So, the change is due to a new good day for GWB, an old good day for Kerry dropping off or some combination of both.

My example (a pure hypothetical) above shows, instead of a one day surge for Bush, a sustained three-day gain for W.

[El Barto]

I considered the point about which day is being dropped from the rolling average.

Today's poll dropped Sept 13 (47.2) which had previouly replaced Sept 10 (47.8) which replaced Sept 7 (47.3). These are not huge swings, so it is unlikely that the Sept 13 one-day result was an unusual out-of-line number.

[GWB]

Hey, they only poll about 300 people a day. This is way too small a sample size.

[b]Over-Confidence Kills. Keep Expectations Low. /images/graemlins/smirk.gif

[El Barto]

[ QUOTE ]
Hey, they only poll about 300 people a day. This is way too small a sample size.

Over-Confidence Kills. Keep Expectations Low. /images/graemlins/smirk.gif

[/ QUOTE ]

You are wrong about the daily number of people interviewed. From Rasmussen:[b]


Every night, Rasmussen Reports interviews 1,000 Likely Voters across the nation to create a three-day rolling average based upon 3,000 Likely Voters.