UPDATE: This diary used to point to discrepancies in Rasmussen's data for Bush. Those disrepancies are not as big as first thought, so I changed the diary (including deleting the overblown rhetoric).
Now I'm posting, instead, estimated 1-day Bush/Kerry samples for
Rasmussen. These estimates are optimized so as to minimize the sum of the mean squared error of the implied 3-day averages and the mean squared error of the implied 7-day averages.
Bush 1d est
Kerry 1d est
44.06745736
48.03111643
48.22925953
45.36040487
49.69088082
44.1735395
49.1839111
46.66333842
49.7654202
45.65901799
49.83258673
45.49521376
45.68655161
47.12063921
47.9472668
47.54006496
48.02841407
46.07102563
47.99512059
47.20976533
49.65404059
45.12436034
49.35516896
43.81063986
46.8943837
47.69476883
49.77198218
44.4911779
48.96794456
45.94295782
48.51046334
45.4206522
46.19090445
47.627069
48.98013259
45.07774657
46.37576741
46.74960308
48.29791854
47.1837525
47.86680531
45.37095161
50.84365647
43.63180803
47.71036448
45.45483518
46.55191969
47.23172152
50.00375775
43.16030616
46.88512047
47.99630121
51.77083529
42.93886242
49.15637874
44.48100438
47.31731072
46.51212455
45.86827546
47.91133675
48.44523451
44.86994138
47.56605817
46.18716619
49.18457055
44.32491228
45.79180936
47.74782716
48.30202695
46.56420167
48.28854991
46.1141474
47.97072266
46.80275043
45.51417849
49.02274776
49.2425307
43.75965762
48.05042655
46.38225328
49.9025843
43.9730664
The top row shows the Bush and Kerry 1-day estimates for 12 October (that is, in reality, the sample obtained by phone on the night of 11th October). The bottom row is the estimate for 2nd September.
With the above estimates, I get the following
average squared errors accross all 3-day plus 7-day
data points:
3-day
7-day
Bush Kerry
Bush
Kerry
.003913538 .001995519
.010086163 .003835394
sum
Bush Kerry
.013999701
.005830913
For reference, here's (some of) the Rasmussen 3-day and 7-day data with respect to which the estimates are optimized:
3-day 7-day
Bush Kerry Bush Kerry
12-Oct 47.4 45.8 47.9 46.2
11-Oct 49.0 45.4 48.7 46.0
10-Oct 49.5 45.5 48.7 46.1
9-Oct 49.6 45.9 48.5 46.5
8-Oct 48.4 46.1 48.4 46.3
7-Oct 47.8 46.7 48.3 46.1
6-Oct 47.2 46.9 48.0 46.3
5-Oct 47.9 47.0 48.5 46.0
4-Oct 48.6 46.1 48.6 45.8
3-Oct 49.0 45.4 48.8 45.6
2-Oct 48.6 45.6 48.5 45.7
1-Oct 48.7 45.3 48.4 45.7
30-Sep 48.5 46.1 48.0 46.1
29-Sep 49.1 45.3 48.1 46.1
28-Sep 47.9 46.3 47.9 46.2
27-Sep 47.8 46.1 48.2 45.9
26-Sep 47.2 46.5 48.1 45.8
25-Sep 47.9 46.3 48.1 45.8
I had a problem with the fact that the mean squared error for the Bush 7-day data is rather high, and over-interpreted it as "fishy business" on the part of Rasmussen. My feet are back on the ground now. :)