11.25.2010 07:58 AM

Six points?

Um, sorry, CBC web heads, but six points is not “significant.”

Margin of error and all that.

Nothing has changed. Nobody has a clear lead. You’re welcome.


  1. Marc L says:

    Warren, you are wrong on this. The margin of error is 2.2 percentage points. So, statistically, at the 95% level, 6 percentage points is significant, which means that the hypothesis that the two parties are tied is rejected.

    • Warren says:

      I’m wrong, am I?

      How many campaigns have you run, Marc?

      • Marc L says:

        That’s not the point. If you’re talking about your ability to close the gap during the campaign, you may well be right. But you’re referring to the “margin of error” in the poll. Well, given the margin of error, the gap is statistically significant. That’s the sense in which your statement is incorrect. Running campaigns and making polls lie is a different issue.

        • Namesake says:

          Technically you’re right, Marc, that there is a statistically significant diff. b/w the two parties at this moment but the problem is the CBC’s pretty much equivocating b/w the two quite diff. senses of “significant” (i.e., a big, meaningful diff. vs. a statistically sig. diff. [which could still be quite small as long as the sample’s big enuff]) in its lede,

          “The federal Conservatives have again taken a significant lead over the Liberals.”

          The thing is — and this, I take it, is WK’s point — despite their implication that there’s been a sig. surge in their popularity, in reality, there may have been NO change in popularity over the past few weeks:

          the Libs are just half a percent off from where they were, after all, and

          even tho’ the Cons are up 4, that’s well within the 2.2 X 2 point magin of error of BOTH polls combined.

          i.e., when it was reported a couple weeks ago that they were at 29.4 +/- 2.2, that might have been an underestimate (as the other polls at the time were suggesting) & they might ‘really’ have been at 31.6, the very top end of the margin of error then (if the _whole_ pop. had been polled); and,

          when it’s reported that they’re at 33.3 now… well, that might be an overestimate, and they might _really_ be as low as 31.1, now.

          • Marc L says:


            What I’m saying is simple. The result of THIS poll is NOT a statistical tie. You (and Warren) can see it as tie in your world, you can consider that it points to a tie in conjunction withother polls, you can see it as an outlier that will be reversed with future polls, you can consider it to be a lousy poll — whatever your judgement tells you. That’s fine, and it’s debatable. But the result of this poll is not a statistical tie, period. That’s not debatable.

            Also, you can’t multiply margins of error on two polls, nor can you consider them sequentially (ie, since the last poll did not show a statistically significant difference between the two parties and this poll does not show a startistically significant difference with the previous poll, there is therefore still no difference between the two parties).

        • Namesake says:

          But now you’re distorting what I said: _I_ didn’t say there was a tie: I conceded at the outset your point that there is stat’ly sig. diff. b/w the two. What I disputed was whether there was a sig. INCREASE in the Cons’ lead in any meaningful sense.

  2. MW says:

    I think you’re right. However, those few of us who pay any attention to polls are getting bored by the fact that nothing has changed in five years. Those who sell papers and want hit on their websites realize this and try to get us to read by pretending there is a story, hence a “significant lead”. Just watch, within the next five days another poll will come out that shows the two parties are statistically tied.

  3. Warren says:

    That’s when I worked for the federal Liberals. I don’t now, and I have no relationship with them whatsoever. I’ve been quite critical of them on Afghanistan, as you know.

    But that doesn’t make significant what is more or less a statistical ti. Nothing has changed. Until the campaign, nothing will.

    • SteveV says:

      “Nothing has changed. Until the campaign, nothing will.”

      Exactly, so just get on with it already. People waiting for “traction”, Ignatieff to catch fire, the Libs to take a nice lead are dreaming. Only an election will train the public’s gaze, until then we chase minor fluctations, within a very well established range that never changes.

      I’d add, does one have to work for the Liberals to be a Liberal? I hope you remember that come the next campaign 🙂

    • JH says:

      Well you guys are all more experienced than I am with regards to polls. However, it wouldn’t matter which party I supported, I’d rather be up 6 than tied or below by 6. If it were the Libs they’d be hooting and hollering about it. Not that I think it is greatly significant – except possibly with regards to MI’s leadership. It and a Fantino victory could cook his goose – this time next year he may very well be having a thanksgiving one back at Harvard. If I were the Tories I’d not be too happy about this poll for that reason. He seems to be their greatest asset at the moment.

  4. Rob C says:

    This is about where the numbers seem to settle back into. Check the EKOs numbers before the last election. No change really.

    Whats frustrating to me is that the constant every time Harper misreads the public and screws up (or maybe he does this on purpose sometimes), we have an LPC that is incapable of building any momentum with voters. The voting coalition needed to defeat Harper has not warmed up to Ignatieff and likely won’t. What’s the answer?

  5. Peter L says:

    Warren, this poll does not suggest a “more or less statistical tie”. Sorry, it just doesn’t.

  6. Michael S says:

    I’m a mathie. “Statistically significant” means “it doesn’t mean absolute shite. It doesn’t mean anything else. For example, “statistically significant” coming out of the quants was construed by Wall Street egos to mean “EVERYTHING IS GREAT”, when in fact it meant “the current numbers mean things are currently going up given the data you give us”. As long as the quants were getting seven figure bonuses for not explaining the details. Those that did got fired. Ask me how I know this.

Leave a Reply

Your email address will not be published.