Has anyone else noticed that many in the media have developed the curious habit of declaring opinion polls within 5% to be tied? When the "Guergisized" Tories are ahead by 3%, the polls are tied; but I guarantee you if the Liberals squeak out even a 0.1% advantage, Jane Taber will write that the LPC have surged into the lead. If a party has been ahead for 10 consecutive polls, even if only by 2%, then it isn't tied. Last night the Phoenix Jets narrowly defeated my Detroit Red Wings, but you know what, let's call it a tie.
An opinion poll attempts to estimate the real proportions of voter intention. Each estimate has a margin of error, which we also call a "confidence interval". These are assumed to have a normal distribution (a bell curve). If you have the Tories estimated at 32% with a 3% margin of error, the probability of the real parameter value being 29% is substantially smaller than it being 31%. A "statistical tie" as you call it should only occur if the probability of it being tied is greater than or equal to the probability that leader is ahead. When you see a 32% to 29% lead, the probability that true popular support is tied is at most 20%, with an 80%+ chance that one party in fact leads the other. Remember GI Jane, confidence intervals are shaped like bell curves, and just because the outer boundaries overlap, that does not imply a tie. Granted I would not be shocked if EKOS had higher than normal error rates due to their history of online polling.
EKOS kingpin Frank Graves is now suggesting based on one poll that there is momentum on Guergis-gate and Detainee-gate. “The resignation and controversy around Helena Guergis and her husband seems to have jumpstarted a moribund electorate in a way that ideas and debate could not.” That is an interesting opinion Mr. Graves, especially considering Ipsos had the Tories up by 10% on Sunday. One poll does not a "moribund" trend make; and if Helena is cleared of any wrong doing, there will be hell to pay with all the false accusations that have been made by the Liberals. And when I see the Green Party over 11%, I know your poll is tainted.
And you know what else; I hardly think a controversy qualifies as "moribund" and a threat to our democracy if Rosemary Barton can't wipe the smile off her face while she's reporting it. If it were as bad as all that, you'd think the Parliamentary journalists would be taking it seriously instead of gleefully indulging in a punditry orgy...
Oh, and Peter MacKay was outstanding in Question Period today! I would like to request that Ujjal Dosangh be tested for rabies. Call it a hunch.
Do you support mandatory breathalyzers before votes in the House?
ReplyDeleteActually, that would be quite funny.
Can you imagine if a confidence vote passed because McCallum was too drunk to be considered capable of voting?
Mind you, he doesn't even know what kind of car he drives on a clear day so how can he ever be considered competent?
Totally agree with you on Peter MacKay. Outstanding he was.
ReplyDeleteBut admit it, you only said "orgy" because Jack mentioned "Eyes Wide Shut" in question period.
In referring to the provincial breakdown, Taber's Tie arguement really breaks down. Refer to http://threehundredeight.blogspot.com/, for a graphical representation, and understandably that the sample sizes get smaller. But if this poll gives a tie to the Liberals,
ReplyDeleteIn Atlantic Canada:
The CPC is at 37% and the Liberals are at 36%. This region, in even the most favorable CPC national polls, tend to put the Liberals in the lead or the NDP-CPC and Libs in a 3 way tie. SO according to Tabers logic, the Conservative will do well and win a plurality of seat in Atlantic Canada, which under the most favorable situation is limited.
More innuendo, more smears in Canadian parliament but only under protection of parliamentary privilege
ReplyDeleteI think the following story is going under the radar but once again illustrates that media are prepared to fly with innuendo and second hand information and smear possible innocent people and is indicative of where politics is going in this country and it doesn't sound like a democratic ,civilized country let alone parliament.
Translator alleges soldiers tried to cover up killing
14/04/2010 7:24:36 PM
The Canadian Press
OTTAWA — A former army translator shocked a parliamentary committee Wednesday with explosive allegations against Canadian troops -- including that they killed an innocent Afghan teen and tried to cover it up.
But Malgarai Ahmadshah, former language and cultural adviser to the commander of Canada's Joint Task Force Afghanistan unit, acknowledged that much of his information was second hand and he couldn't offer any proof to back up his claims.
He also wouldn't repeat his allegations outside the committee room, where he was no longer protected by parliamentary privilege.
Inside the room, Ahmadshah claimed Canadian soldiers shot a 17-year-old boy in the back of the head in a village north of Kandahar in the summer of 2007. He said they shot the teen because they thought he had a gun.
He also alleged the soldiers tried to cover up the alleged murder by rounding up men in the village.
"After the Canadian Forces wrongly killed a man, they panicked. They swept through the neighbourhood, arresting people for no reason," he said.
"They arrested 10 men, from about 10 to 90 years old. All the men were taken to Kandahar airfield, where I personally interviewed them with a military tactical-questioner unit.
"None of the detainees were Taliban. None did anything wrong except to be at home when the Canadian Forces murdered their neighbour. Yet Canada transferred all these innocent men to the NDS. I don't know what happened to them."
Ahmadshah later told the committee that he wasn't in the village when the teen was allegedly killed. He said his knowledge of the incident came from interviews with detainees and military documentation.
The Afghan-Canadian man spent 13 months with the military in Kandahar in 2007 and 2008. He was known as "Pasha" to soldiers and Afghans when he served Brig.-Gen. Guy Laroche, commanding officer of Canada's military effort in Kandahar at the time.
Ahmadshah testified that Laroche and other top military brass knew that prisoners handed over to Afghanistan's notorious secret police would be tortured.
"There was no one in the Canadian military with a uniform who was involved in any way, or at any level, with the detainee transfer that they did not know what was going on and what NDS does to the detainees," he said.
He said his military bosses ignored his concerns every time he raised them.
That triggered a fiery exchange with Conservative MP Laurie Hawn, a former air force colonel. Hawn asked if Ahmadshah was calling a succession of Canadian generals liars.
"I don't call nobody a liar," Ahmadshah said.
http://news.sympatico.ctv.ca/Canada/ContentPosting?newsitemid=CTVNews%2F20100414%2Fafghan_detainees_100414&feedname=CTV-NATIONAL_V3&show=False&number=0&showbyline=True&subtitle=&detect=&abc=abc&date=True
yah, we'll see who gets "Guergisized"
ReplyDeleteThe backbone of Canadians is not socialized medicine Iffy ,
the backbone of Canadians is fairness.
That's why the Canadian revolt over the revolting coalition of losers, it was not fair.
And we will see what Canadians think of this latest smear job by the collective media and opposition parties.
How is it MPs use immunity in the House but won't say a word 'outside' without protection,
but media broadcasts 'protected' proceedings, live?
I see an 'opportunist'.
ReplyDeleteWhy does the Canadian media take this man at his word?
The margin of error in the EKOS poll is 2.5% 19 times out of 20, meaning that is a 95% confidence interval. That is to say, 95% certain that the true parameter value is within 2.5% (but the further you get from the estimated value, the less likely it is). The Tory lead in that EKOS poll is 2.4%. The bell curves will overlap on the tails, meaning the probability that the true parameter values are tied (or LPC ahead) is comparatively small. Do I need to draw you a picture? If the probability it is tied is less than 30%, can you consider it statistically tied?
ReplyDeleteAnonymous, if you'd like to have a mathematical discussion in my comment section, drop the insults. I did study probability distributions as part of my Mathematical Economics degree.
Ted Menzies was also great in QP today, especially when he asked the Honorable member if he could tell were the 57 million stolen from the EI fund was. If it hadn't been stolen perhaps rates would not have to be raised.
ReplyDelete