I have been getting it in the neck from a patronising commenter for making a statistical error.
I divided an average figure into a median figure. Wrong.
The corrected figure is;
Government transfers account for 8 percent of the average male income and 19 percent of the average female income (not 8 percent of the median male income and 24 percent of the median female income, which I originally posted.)
The commenter goes on the tell me what I should have done;
Taking the median (if you want to use medians) weekly govt transfers by gender from Table 12 and putting them against median aggregate income in table 1, you'd be horrified to know that the median male weekly govt transfer income is 40% of the median male income from all sources, but the ratio for females is 64%.
So why the huge difference between the two sets of figures? My erroneous figures looked far more likely than these, which is why I didn't spot the mistake.
Because Table 12 is Median Weekly Income by Income Source for those receiving that source of income.
Hence it is an error to divide those figures into the figures at Table 1 which is Average and Median weekly Income for all people aged 15 and over.
So the anonymous commenter, who demands I admit my mistake or lose my credibility, might like to join me. Especially as he or she says "what really irritates me is the abuse of statistics."
2 comments:
Lindsay I would give no credence to any anonymous poster.
If they don't have the nous and courtesy to at least use a psuedonym I wonder how they even have the brains to work out you made an error.
Ha - good call. I obviously fixated on "median by income source" and went forward from there. As I also said, everybody does it. Probably need a holiday.
Anyway, between the two of us we've demonstrated that the dataset in question is insufficient for the purpose at hand. Fine - now you can actually look for data that is sufficient to prove your point. Median income by source, all people receiving an income?
Another option might be distribution curves for amount of income from the DPB (OIA work & income??) Not having seen that data I would be interested in the result, particularly if it were matched by a distribution controlled for by number of dependents of the recipient.
Real-world evidence that doesn't logically support a belief might be enough for you to take it on faith, but it is no reason for anybody else to believe it . If you want to actually inform or persuade people, as opposed to being just another opinionated blogger who wants attention, get the data to back it up.
As for the matter of why I choose to post anonymously, the only reason that would inhibit the discussion is if you wanted to play the player and not the ball. I might have a good reason, or I might just be a coward, or I might just prefer to have no linkage between this discussion and my work or home life.
As you have discovered, the evidence and sources should be enough, and are neutral - data doesn't care who's position it supports (if anybody's). I note that criticism of anonymous comments for being anonymous has not been made of commentators who agree with you, or the ones who suggest "9mm" solutions to poverty and crime.
A.
Post a Comment