Recently, I came across this particular article in ToI. After reading the article, I chanced upon the comments section. I was quite surprised to see that most of the readers have accused Dr. Venkatraman Ramakrishnan of being an idiot, moron, a terrorist, not respecting Indian culture and quite a few have gone berserk with demands to withdraw Nobel Prize. A couple of comments show surprise by the fact that he doesn’t believe in astrology inspite of being a Nobel Laureate in Signs. Fine, I made up the last line. 😛

Anyway the main question is how do you test for the credibility of something like astrology, existence of God and other abstract things which is militantly supported by more than a billion people? It may be true that astrology lacks scientific evidence but there maybe, just maybe an iota of truth in it. Maybe it is not yet good enough to be called Science. What if the astrological “principles” are sort of pointing in the correct direction but we are not able to grasp or make a breakthrough? The problem is that if a person dare says a word against astrology, the entire populace of India seems to find trolling an enjoyable idea. It is difficult to argue with people who weave stories about how astrology predicted one particular event in the life of one particular person. This is where the beautiful concept of Null Hypothesis can make the difference.

The main idea of the Null Hypothesis is that you assume a certain stance unless proven otherwise. Or to put in more compact terms, the default position. To demonstrate the idea of Null Hypothesis, let me take an example of a random coin toss experiment. I assume that the default position is that the coin is unbiased, that is, there is as much chance of getting heads as there is of getting tails. So, I toss the coin 10 times. Now the expected number of heads that I should get is 5, assuming independent trials. But what if I get 7 heads and 3 tails? Can I declare “confidently” that the coin is biased? Not really. What if I toss the coin 100 times and I get 70 heads? Just maybe. What if I toss the coin 1000 times and I get heads 700 times? Now it is more apparent that the coin is biased, but is it biased (0.65,0.35) or (0.70, 0.30)? I do not know that yet. I need to conduct more experiments. Of course this is just the property of Law of Large Numbers. The question is how many trials do I require before being say “95% confident” that the coin is biased? There is a bit of mathematics involved in it, which I will not go into the details of. So let me assume that in I need 70 trials out of 100 to turn up heads to declare that with 95% confidence the coin is biased. This is termed as “Rejecting the Null Hypothesis with 95% confidence”.

Now that the tutorial on Null hypothesis is over, let’s see what can be done regarding astrology. The main problem with astrology is that there are so many contentions regarding the rules that it becomes difficult to have a comprehensive test. So, a manual regarding the rules of astrology needs to be prepared such that there are no evident contradictions. If the argument is that astrology is too vast a subject to be put down into a manual, let them take a small part of it, say zodiac sign and one dominant characteristic of the person belonging to that sign. Now we need to prove that “most” of the times the Zodiac signs and the trait in question are predictable. So, it becomes a binary random variable akin to the coin toss experiment – heads if the person has the same traits as the one predicted by the Zodiac sign, tails otherwise. One obvious problem is what the Null hypothesis should be. If my thesis is that the theory of zodiac sign is truly random, we take (0.5,0.5) as the null hypothesis. If I believe that it predicts the behaviour 60% of the times, my null hypothesis would be (0.6,0.4).

Either way, it should be an interesting experiment. I wonder about the technical feasibility of such a project. Well, if not the entire nation, maybe at a smaller scale inside the insti campus.

Any views? 🙂