« Double dip risks | Main | Jail Time and the Census »


Feed You can follow this conversation by subscribing to the comment feed for this post.

The more that various Conservative MPs get roped into the census debate, the more tangled this web becomes. Your post, along with Bernier's quotes raise the latest question, in a growing line of questions.

This is the question: Are MPs, like Maxime Bernier a) fabricating evidence to help back up Harper's decision to axe the census long form (eg, the claim about receiving 1000+ emails per day on the matter in 2006); or b) telling the truth about having received 1000+ emails/day in 2006, thereby demonstrating gross incompetence and unaccountability for having never brought the issue forward (as sources, such as the Hansard record indicate).

So, which is it, Conservative government deception or Conservative government incompetence?

Another fun search to do for journalists and academics: Do a LexisNexis, Proquest or Factiva search for Bernier+census and see what you come up with. Answer - not a whole lot before this weekend.

More fun - I did a Lexis Nexis search for "Canada + census + privacy" for articles written in the year 2006. A few articles came up about changes to the form which "asks for permission to publicly release personal census data 92 years from now."

Other than that, I found *2* entries. Both of which were in the form of letters to the editor, both of which to the Toronto Sun. July 2, 2006:


My family and I were enjoying a Saturday morning at home when we got an uninvited guest knocking at our door. We were not expecting anyone and we were trying to have a quality time together. So, we did not pay attention and thought that perhaps it was someone selling something or Jehovah's Witnesses trying to win converts. The person at the door continued knocking. Our unexpected guest was from Census Canada, who reminded us to fill out the questionnaire that was due May 16. The visitor was nice and friendly, but we got the impression it was not a reminder but a warning -- fill out the forms or else. With all due respect to our government, I believe this census is nothing but a waste of our taxes and resources. It will not help resolve anything. It is a waste of resources that should be used in different ways to help build our country and help those who need a helping hand. The government should not intrude and invade people's privacy by forcing them to complete such meaningless forms.



(Too bad, it's the law -- and it's not an invasion of privacy. Governments have been doing a census of some sort since Biblical times)"

and July 9, 2006:


Re "Census is senseless" (Letters, July 2): Canada's first census was initiated by Intendant Jean Talon in 1666. We have travelled 335 years of census history in Canada. We have grown from 3,215 inhabitants to a nation of almost 31 million. Census-taking in Canada has adapted and changed as we have grown from a colony into a country comprised of ten provinces and three territories. Statistics Canada conducted the Census to develop a statistical portrait of Canada and its people. The census is here to allow us to stand and be counted, to be a part of Canada. To say you are one of those citizens. The only tax dollar wasted is when you do not fill out the forms and send them in. Unless you have something to hide what is the big deal? You are Canadian, are you not -- do the census. Be Canadian and be counted.



(Gripes about the census being an invasion of privacy are sadly misguded)"

That's it. If there was such a big uproar about privacy concerns in 2006, why didn't the media pick up on it?

Here is how to solve the privacy problem: Cancel the long form all together, and on each individual short form randomly assign between 0 and 10 questions from the long form.

It would take me about a week to write the software and document macros to accomplish this on an industrial printer.

Bonus - you can use your choice of randomized block design!

Aaron - how, though, do you correlate between the questions? I think that is part of the purpose - not just getting a bunch of independent data points.

Through marginalization, good random block design, and by recognizing that not all contingencies are relevant; in short by carefully formulating hypotheses on contingency before constructing the randomization of the survey questions. The sample size is large enough to be able to test any of the suspected contingencies between variables, even through random assignment of questions; provided you propose the contingencies a priori, and design the randomization around testing those contingencies.

Quite frankly, even if every one of the 50 questions was binary, you still have a sample space of 2^50, which can never be covered by 35000000 people, so fishing for contingencies between questions one will eventually always find them (there is an equivalent problem in genome wide association studies). A random assignment design will actual improve the experiment, by forcing a careful choice in the blocking of questions together to provide appropriate coverage of the sample space.

Can anyone find anything from 2006 RE: privacy concerns to the census that is more substantial than a Toronto Sun letter to the editor, which the Sun dismissed as being "not an invasion of privacy"?

Here is an initial reference for random design http://www.jstor.org/stable/2291129

One can also search for Multiple Matrix Design by Shoemaker, 1973

Some quick design math:

If there are 50 questions on the long form and they are divided into 10 blocks of 50/10=5 questions, then every pair of questions could be asked with only 10*9/2=45 different short census versions, adding only 10 additional "intrusive" questions to the short form.

Finally each pairing would have a sample size of 35000000/45=777777

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad