Saturday, July 3, 2010

Space in MRT trains

Hi Mr. Tan
Recently there is a debate going on whether the SMRT trains in Singapore are overly crowded or not. The SMRT doesn't think very much of it as evidenced in their CEO's untactful remarks, whilst practically everyone blogging on the internet curse and swear at how bad the situation had become.

Looking at the case again, I just discovered that everything hinges on a statistic which is provided by SMRT Quote:...This averages 3.8 passengers per sq m, lower than that of major cities like London..... Unquote.

So 3.8 passengere per sq m is the agreed benchmark. I have no problem with that.  But look at the word AVERAGE. Therein lies the solution to the puzzle. It's not about the number 3.8 neither is it about the techie word  crushload. It's about AVERAGE.

AVERAGE is a very dangerous concept if not properly managed. If the annual salary of the PM is $3 million and the annual salary of the rest of the 9 ordinary workers in the PMO (PrimeMinister's Office) is $50,000 then mathematically the average annual salary of the people in the PMO is $305,000 per person? In statistical theory there are median, and mode, standard deviation, etc to refine the imperfect concept of averaging. In the real world, we also need to know more about the model used for "averaging".

How does SMRT calculate its AVERAGE crush load? Is it over time average between peak period and non peak period per station basis? How many stations are selected? Which are the stations selected for the averaging?  Are newer stations which have lower load factor bundled together for averaging?

Statistics is a powerful tool which can often lead to wrong conclusions. Sometimes it is so powerful that it can be used as a tool to make one feel complacent and look good. And those who ignore the possibly distorted conclusions might be told to make other choices because the sysem is just doing fine - on "average". 

rex