I’m swearing right now.
A lot.
I have just found an error in my formula for calculating the number of seats at the low end of a poll’s margin of error. A really subtle error, too, which is why I missed it before today.
I had a closing parenthesis out of place by one term in a formula, which, for the NDP, threw absolutely everything on the low end off.
So I’m going to re-post a selection of corrected predictions for every poll since April 24.
The first number is the seat count predicted based upon the published results of the poll. The numbers in parentheses are the two extremes, based on the poll’s published margin of error.
April 25
Nanos:
Conservative: 166 (160-170)
Liberal: 50 (50-53)
NDP: 45 (40-48)
Bloc: 47 (47-47)
Ekos:
Conservative: 144 (143-144)
NDP: 107 (107-108)
Liberal: 43 (43-43)
Bloc: 14 (14-14)
April 26
Angus Reid
Conservative: 147 (146-147)
NDP: 100 (99-100)
Liberal: 44 (44-46)
Bloc: 17 (17-17)
April 27
Nanos
Conservative: 162 (159-162)
NDP: 92 (92-94)
Liberal: 38 (38-39)
Bloc: 16 (16-16)
April 28
Ekos
Conservative: 147 (146-147)
NDP: 98 (98-99)
Liberal: 48 (48-48)
Bloc: 15 (15-15)
April 29
Ipsos Reid
Conservative: 152 (152-157)
NDP: 124 (119-124)
Liberal: 25 (25-25)
Bloc: 7 (7-7)
As you can probably guess, I’m not very happy about finding an error in my formula at this late point in the campaign, however I think it’s better to disclose it and correct my predictions.
Yeah, your previous ranges looked… odd. But the ones you have now seem awfully narrow…
There are some assumptions built into the ranges which may be throwing them off a bit. Who knows? At this point, I don’t intend to make any adjustments as to do so would mean a LOT of new calculations and formulas; and there really just isn’t time.