Wednesday, February 13, 2008

Income in Singapore

Singapore newspaper likes to publish articles that talks about statistics released by government.

I myself am rather fed up with statistics reporting because I know personally how statistics can be easily "played" to produce figure to your liking.

Outliers for example is always easily cleaned out since they are said not to "represent" the populations. Simply said, the extremist are not to any statistician's liking. They don't care that the people called "outlier" may actually constitute the "hate group" which may bring down the statistics (e.g. people who proclaim that product A is sucks and may some times carry the truth) or the "love group" which may help to shore up the validity of the statistics (e.g. people who love the product so much that they spend bulk of their income to consume or promote it). There is of course chance for cleaning one side of the outlier to skew the result. Who will audit the result anyway?

Second of all, I realize what's always missing in the newspaper's statistics value is standard deviation or even the range. Saying that average is X makes no strong argument if the standard deviation is huge or constitute as big as the X. If average income is S$6,830 but the range of the sample is S$100 - S$100,000, do you really want to believe at the average value? Will average of S$6,830 when 70% of the population is at the range of S$100 - $ 6,000 destroy the trust of the statistics? Make your own judgement. That's simply why range or standard deviation is rarely stated. It simply reveals too much of the truth.

Thirdly, statistics value rarely tells you whether it's a normal distribution or a skewed distribution. After all, if the second point above happens, the statistics skew is actually skewing to the left. What does it tell you if the bulk population is actually far poorer than the average value?

Lastly, surveyors can always choose who they want to be in the statistics sample. Who will prevent them from choosing the high end income group for sampling? Will the reader know it? Nah, the surveyors and the statisticians are as safe as Singapore.

No comments: