Two comments on the ongoing censusgate saga:
- The Chief Statistician has resigned his position in light of the dispute: Fantastic, good for him. He should have done it sooner but it seems last night’s ‘clarifying’ interview with Minister Clement pushed him over the edge. The Minister pointed out that some people at StatsCan believe the agency is arms length, but in fact it is merely a tool of their governing minister, ie. himself. Maybe it should be an arms length agency? How about that for an NDP or Liberal campaign promise?
- Mr. Clement had some real nuggets in that interview, I’ll just mention the one pointed out on Challenging the Commonplace. The Minister tweetith:
168,000 felt strongly enough last time about mand long form to refuse on pain of jail. Yet that sample was deemed valid.
That’s an interesting factoid, we’ll need to look into it. But consider that 168,000 out of the roughly 6.6 million people the survey was given out to (that’s 20% of 33 million people) makes it about 2% of the sample. Even if it was given out to a lot less people and it was 10% of the sample it would be bad, but not nearly as bad as what they are doing now. Losing a few percent out of a huge sample set isn’t as damaging to the results as skewing 100% the sample with noise that cannot be controlled for. We simply can’t know who will choose to fill out the survey and who won’t, so it won’t represent the actual population. To work it out you’d need extra polls and studies to correct for it, something the US already tried and abandoned because so few people filled out the voluntary census.
So, nice try Mr. Clement, what’s your next idea?