How many samples are enough to build the Kidex highway?


There’s a highway they want to build from Damansara to Puchong–called Kidex, and just like any other highway before, people are understandably worried about the construction. This excerpt from the KL-Chronicle details the causes of anxiety:

[box icon=”chat”]Kidex will be constructed over heavily built-up residential areas in Petaling Jaya and will pass very close to schools, houses and places of worship. It will pass just 5m away from two schools – Bukit Bintang Boys Secondary School and Sri Petaling Primary School. Its distance from the Tun Abdul Aziz Mosque in Section 14 is listed as 7m and from St Paul’s Church as 18m. Houses in parts of Sections 2, 4, 7 and 8 will be just 10m from the highway

And so, when Kidex had their townhall last week, a group of protesters showed up to voice their displeasure–as should be allowed in a democratic society. Kidex claimed they had conducted a survey that proved that the majority of the people wanted the highway, this was hotly contested by the Say No to Kidex committee, who contended that the survey wasn’t ‘authentic’.

The video below (from Malaysiakini) has a great interview with the secretary of the Say No to Kidex committee outlining their points of contention on the survey by Kidex. (starts at 1:10)

Here’s the seven points raised by the Say No to Kidex committee.

1. The initial Kidex survey of 300 respondents–of which 73.4% were agreeable to the building of the highway.
2. The Say No to Kidex committee did their own survey on 20 different locations, including the Mosque, Schools, and the resident associations of the areas affected by the highway.
3. Say No to Kidex can’t comment on how many people they’ve engaged but can confirm it was more than 300.
4.Hence the public perception of the initial Kidex survey is negative.
5.The list of the initial 300 respondents has not been shared–as it was confidential.
6.The survey was conducted by a Ph.D in statistics, but this survey itself was funded by Kidex.
7.The next survey as planned by Kidex would have 2000 respondents.

Now let’s take this apart one by one, because there is some maths here is quite foreign to most–this is the world of probability and statistics. 

Point 1: The first Kidex survey had just 300 respondents

This is a valid point from Kidex. They did the survey and the results were quite good, unfortunately that’s only half the story. What’s important is we know the confidence interval and confidence level of the survey. Surprisingly, even if PJ had 2.4 Million people, to achieve a confidence level of 95% with a confidence interval of 5–you only need 385 respondents. Basically that means you should be 95% confident that the results is within +/- 5%

To the layman, what that means is 300-400 people would be enough to gauge public perception, particularly if the results of the survey were so heavily in favour of the Highway(~75%). For example if you surveyed just 100 people, but 99 of them agreed with the highway, your confidence of the result would be much higher than if you surveyed 100 people but only 50 agreed.

So 300 is a good enough number for an initial survey.

The better question would be “Who were the respondents?”. If the 300 respondents were just the family and friends of Kidex employees, that wouldn’t make for a very good survey with even 1000 respondents. Which brings us to the next point.

Point 2: Say No to Kidex has engaged the relevant stakeholders

Yes and No. Obviously asking the people directly impacted by the highway would result in a resounding NO. However, at least that’s a start. What’s relevant is also to ask those that don’t stay in PJ, but work there and would use the highway. I don’t stay anywhere near the LDP, but use it regularly to travel to Cyberjaya.

At least Say No to Kidex can point out their stakeholders–Kidex hasn’t.

Point 3: They surveyed more than 300

Unfortunately, this is a red-herring. 300 is good enough, and 600 is better. But going to the thousands doesn’t really help in the sampling. In fact, you reach a point of diminishing returns, and so the number of people sampled in the survey is less important than how the respondents of the survey were chosen.

The more relevant point to be made, is that the results of both surveys are so widely different that the accuracy of both surveys should be brought to question. Which is where it can be rightly said that “the authenticity of the survey is questionable”.

Point 4: Hence the public perception of the initial Kidex survey is negative

This is absolute nonsense–the survey is built to gauge public perception. If you think that there’s a negative perception–run another survey to prove it. You can’t dispute science with perception. If you want to disprove a survey result, point a flaw in the design–not just throw about public perception because it shouldn’t count.

The perception of zombie apocalypse is at all time high–but we’re not doing anything about it now are we?

Point 5: The list of the initial 300 respondents has not been shared–as it was confidential

OK la. The list is confidential

But it would be interesting to see the demographics. Where did they live, where did they work–how much would they pay for the highway–etc etc. This again leads to the question of who the respondents were.

Point 6:The survey was conducted by a Ph.D in statistics, but this survey itself was funded by Kidex

This is NOT a Red-Herring.

Studies show conclusively that the company who pays for the surveys usually get the result they want. Quoting the wikipedia article on the Bad Pharma:

[box icon=”chat”]In “Missing Data,” Goldacre argues that the clinical trials undertaken by drug companies routinely reach conclusions favourable to the company. For example, in 2007 researchers studied every published trial on statins, drugs prescribed to reduce cholesterol levels. In the 192 trials they looked at, industry-funded trials were 20 times more likely to produce results that favoured the drug.

So to suggest that the good Doctor may have been biased is with merit. It doesn’t cast judgement on his character but rather points out conclusive studies to suggest that Kidex shouldn’t be funding any survey to gauge the perception of their own highway.

Point 7: The next survey from Kidex would have 2000 respondents

What’s the point right, we’ve already seen that anything more than 700 would be redundant. In fact 700 respondents would give you a confidence interval of 5, with a confidence level of 99% (given a population of 2.4Million)

The important thing is to ensure the RIGHT respondents are chosen, with the right mix of stakeholders.

That’s where the discussion should be–and not trying to get more respondents.


So hopefully that clears some things up. These are tough things to discuss, because peoples homes are at stake, but then again the benefit for society in general could be enormous. The important thing is that we fully understand the problem and the impact before we proceed, and for now there seems to be little information on both.

Add comment

Astound us with your intelligence