Tuesday 29 May 2012

Statistics at Work


Your problems? Solved.
email nick leth at gmail dot com. No worries. Now.

Statistics, eh! Marvellous things, statistics. Natural sow's-ear-to-silk-purse converters. For both researchers and reporters...

Aussies waste time at work according to Ernst & Young researchers. I say "researchers" to be polite. They analysed a survey. Perhaps the problem is with Ronan O'Connell, who reported on the survey (The West, 28 May 2012).

According to the report, "one third of Australia's workforce did not meet the national 'productivity average'."

Goodness me! One third of workers are below average productivity! So how is this "average" computed?

One third of Australians are below average height! One third of Australians are below average weight! One third of Australians are below average age! Wow!

An "average" will always have some above and some below. That's why it's called an "average".

If the research calculated a "median" productivity then exactly half of Australian workers would be below that median value. How shocking is that?! Not shocking at all, because of the way that a median value is calculated.

One third of Australian workers work at below this mysterious "national productivity average"... So what?! Two thirds work above... The average is -- by definition -- somewhere in the middle ... at an average value!

If the survey report provided an independently calculated "national productivity expectation" -- that would be interesting. But an "average" -- presumably based on the workers who were surveyed -- tells us nothing.

Nothing at all.

Except about the quality of the survey. Or the quality of the survey analysts. Or the quality of the newspaper reporter.


Independent thinking & independent analysis of your problems.
Agamedes Consulting: Support for your thought.
email nick leth at gmail dot com

1 comment:

Anonymous said...

Nick, mate! You've really missed an opportunity here: " average" can mean mean, mode or median. Most people mean mean, of course. How in (whatever!) can such a raw statistic convey any real meaning? I mean, are these "researchers" having a lend, or what? We can assume a skewed distribution, but little else. How did they measure efficiency? What assumptions were made? Oh dear! Tut-tut!