I'm enjoying a little self-satisfaction today; Torontoist picked up my Ford Nation passport cover last night. So as a result, I'm being a little vain – checking my flickr stats, hit count on the Torontoist page and google analytics on the blog. It's nice; the numbers aren't big, but they're big for me.
One thing I still find funny is that, according to google analytics, people still arrive here by googling for hummingbirds coming from cocoons. My post is hit number 3 when I search for it. Unfortunately, not too many people stay long on the site when they arrive that way, but occasionally they surf around a bit or leave a comment on the post. But even in the sea of people showing up from flickr and Torontoist (and my own facebox sharing), I still see a few hits for hummingbirds coming from cocoons.
Showing posts with label statistics. Show all posts
Showing posts with label statistics. Show all posts
Saturday, March 5, 2011
Thursday, October 14, 2010
Lies, Damned Lies and Opinion Polls
The Toronto Star is trying a new tactic in its war on informed democracy today.
Up until now, opinion polling numbers have been reported as fraction of decided voters. So, for example, when they said that Ford had 39% of voter support, what they meant is that 39% of the 64% of the people polled said they would vote for Ford. Which actually means that 24% of the people polled said they would vote for Ford. This tactic was useful in scaring people into thinking they need to vote strategically – I would be willing to bet that almost none of the undecideds were undecided about whether they were going to vote for Ford.
But this morning, they have cleverly switched to simply reporting that second number. Before, if you added up the percentages for each candidate, you'd arrive at 100% and the undecideds would be extra. But today, if you add up the numbers for Ford, Smitherman, Pantalone and Rossi, you get 76% which is conspicuously close to the number claimed in the article as decided voters.
The Star is using this to show the runaway-freight-train-like momentum of Smitherman's campaign. Of course, if you look at the real numbers, Smitherman's current rating looks a lot like his old rating plus Sarah Thompson's rating. So basically, at this point, all the candidates are within the margin of error of their previous values. In fact, if you look back on the opinion polls, there has been basically no change since July, with the exception of Thompson's dropping out.
So again, if you want to feel all clever because your strategic vote kept the Stupidest Man Who Ever Ran For Mayor* out of office, go ahead. We all know how that ends.
Or, you can do your research, learn about the candidates plans for transit and if you still want to vote for Smitherman, we'll argue some more.
* I realize that there have probably been stupider candidates. It's even possible one of them won.
Up until now, opinion polling numbers have been reported as fraction of decided voters. So, for example, when they said that Ford had 39% of voter support, what they meant is that 39% of the 64% of the people polled said they would vote for Ford. Which actually means that 24% of the people polled said they would vote for Ford. This tactic was useful in scaring people into thinking they need to vote strategically – I would be willing to bet that almost none of the undecideds were undecided about whether they were going to vote for Ford.
But this morning, they have cleverly switched to simply reporting that second number. Before, if you added up the percentages for each candidate, you'd arrive at 100% and the undecideds would be extra. But today, if you add up the numbers for Ford, Smitherman, Pantalone and Rossi, you get 76% which is conspicuously close to the number claimed in the article as decided voters.
The Star is using this to show the runaway-freight-train-like momentum of Smitherman's campaign. Of course, if you look at the real numbers, Smitherman's current rating looks a lot like his old rating plus Sarah Thompson's rating. So basically, at this point, all the candidates are within the margin of error of their previous values. In fact, if you look back on the opinion polls, there has been basically no change since July, with the exception of Thompson's dropping out.
So again, if you want to feel all clever because your strategic vote kept the Stupidest Man Who Ever Ran For Mayor* out of office, go ahead. We all know how that ends.
Or, you can do your research, learn about the candidates plans for transit and if you still want to vote for Smitherman, we'll argue some more.
* I realize that there have probably been stupider candidates. It's even possible one of them won.
Thursday, October 7, 2010
I Can Clearly Not Choose The Candidate In Front Of You!
Yesterday a long-time supporter of Joe Pantalone flopped and is now officially endorsing Smitherman for mayor. Let's forget, for a moment, that the opinion polling numbers are biased, meaningless and poorly reported and just remember that voting "strategically" is a good way to get a bad mayor while feeling really clever and self-satisfied. I think that saying "I think Joe Pantalone would be the best mayor but I'm going to vote for Smitherman" puts in you a similar position to the Sicilian from the Princess Bride. And we all know how that ends:
In any case, I'm with Janet Davis who said “I am strategically voting. I’m strategically voting by endorsing Joe Pantalone, the best candidate for mayor and for the future of our city.”
In any case, I'm with Janet Davis who said “I am strategically voting. I’m strategically voting by endorsing Joe Pantalone, the best candidate for mayor and for the future of our city.”
Friday, July 23, 2010
The Choice of Tyranny
For those of you not following, presumably because you live in some other country than Canada, our government has done something supremely stupid.
Statistics Canada, or statscan, is a world renouned stats collecting body. One of the reasons for this reputation is the mandatory cenus – all Canadians are required by law to fill out the census. As well, one quarter of Canadians are required by law to answer a long-form (53 questions instead of the usual 8) questionnaire. This mandatory thing is important, it is ostensibly the crux of the matter.
You see, the current government has exercised its droight de seigneur and made it no longer mandatory to answer all 53 questions in the event that you draw the short straw, so to speak. Statisticians are, rightly, pissed.
The reason the government claims for doing this is that it is an invasion of privacy. Of course, they also claimed that they were doing this with the consent of the head of statscan who promptly resigned to prove that they were lying.
As someone who has used statscan census data in the past, I can tell you that the privacy argument is ridiculous. As much of the data as possible is released, but there are certain products for certain census tracts which are held back because the small number of responses might allow someone diligent enough to use the data to identify individuals. The statscan website FAQ also points out that not even contractors are allowed to see the data.
So, what is really going on here? In Canada we have a government system which is set up to limit the power of any individual. (This is in stark contrast with our neighbours to the south who elect a new king ever 4 years.) For 4 years now Harper has been stuck with a lousy minority government, which means constantly having to consult with the opposition parties instead of just riding roughshod over the country's laws and ideals. I have been led to believe that the entire cabinet is completely under the thumb of the PM, which means this census decision came directly from Harper himself. And I think it has nothing to do with logic or reason, privacy or the census, even. I think this is entirely a case of Harper doing something because he can do it and the opposition parties can't do a damned thing about it.
Statistics Canada, or statscan, is a world renouned stats collecting body. One of the reasons for this reputation is the mandatory cenus – all Canadians are required by law to fill out the census. As well, one quarter of Canadians are required by law to answer a long-form (53 questions instead of the usual 8) questionnaire. This mandatory thing is important, it is ostensibly the crux of the matter.
You see, the current government has exercised its droight de seigneur and made it no longer mandatory to answer all 53 questions in the event that you draw the short straw, so to speak. Statisticians are, rightly, pissed.
The reason the government claims for doing this is that it is an invasion of privacy. Of course, they also claimed that they were doing this with the consent of the head of statscan who promptly resigned to prove that they were lying.
As someone who has used statscan census data in the past, I can tell you that the privacy argument is ridiculous. As much of the data as possible is released, but there are certain products for certain census tracts which are held back because the small number of responses might allow someone diligent enough to use the data to identify individuals. The statscan website FAQ also points out that not even contractors are allowed to see the data.
So, what is really going on here? In Canada we have a government system which is set up to limit the power of any individual. (This is in stark contrast with our neighbours to the south who elect a new king ever 4 years.) For 4 years now Harper has been stuck with a lousy minority government, which means constantly having to consult with the opposition parties instead of just riding roughshod over the country's laws and ideals. I have been led to believe that the entire cabinet is completely under the thumb of the PM, which means this census decision came directly from Harper himself. And I think it has nothing to do with logic or reason, privacy or the census, even. I think this is entirely a case of Harper doing something because he can do it and the opposition parties can't do a damned thing about it.
Labels:
canada,
census,
government,
harper,
politics,
statistics,
statscan,
tyranny
Subscribe to:
Posts (Atom)