Want to be an Ass? Just Lie With Statistics

Mention the word statistics to any ordinary audience and the faces around won’t disappoint with their myriad contours. Statistics has been reduced to its barest to make it conjure only numbers. Well that could be true but the truth remains there is more to statistics than mere numbers.

Acalculi (general fear on numbers) is the genesis of most people demonizing statistics. For the records there is mere counting and number arrangement (combinatorics) and statistics just like there is number manipulation (arithmetic) and mathematics. So what the heck am I up to with all these? Well, I am not splitting hairs nor engaging is a simple game of semantics.

Anybody engaged in any form of research inevitably becomes a statistician on sorts. Yes that research question of yours is nothing but a NULL HYPOTHESIS. And you do your thing and end up ‘proving’ or ‘disapproving’ it. That last bit is decision premised on ALTERNATIVE HYPOTHESIS. But again why all the heat with these two ordinary words – null and alternative? And there lies the catch, you either get it or miss the boat by a river and your research is everything but DOOMED.

Three brands of liars
A statistician will have the hardest time explaining the ordinary words to anybody outside her profession. As a chartered epidemiologist/statistician I found it one of my greatest challenges explaining to a jury what these MUNDANE terms mean in everyday life. So here I was on standing precariously with my profession at stake before legal scoundrels ready with hammers and six-inch nails to puncture holes into my evidence. All they want from me is an exhaustive and unambiguous explanation of types I and II errors. See already you are numerically blushing at my preference for Roman numerals instead of the ordinary 1 and 2!

Well to cut a long story short, hypothesis testing and estimation are the two singular important wheels in the statistical cog. Hypothesis is nothing but an informed statement about an unknown population parameter whose plausibility we want to evaluate using information obtained from a sample of the same population. Such an assertion becomes the Null hypothesis and it can be either true or false its complementary becomes the ALTERNATIVE hypothesis.

You make a mistake by rejecting the assertion if it is indeed true and consequently commit type I error. On the other hand you make a mistake by failing to reject the assertion if it indeed false hence committing type II error. So which of these errors is greater sin than the other? I gave the jury the analogy of convicting the innocent (type I) and releasing the guilty (type II). The law being the ass it is often designed to be would scream itself hoarse never to contemplate type I error at any cost.

So can you still lie with statistics? Granted there are three types of liars: liars, damn liars and statistics. Whatever brands of lies you prefer never make an ass out of yourself.