The Sample With The Built-in Bias

Authored By: Isabel Anil Joseph (M. Sc. Statistics and Data Science)

Do you swear to tell the truth, the whole truth and nothing but the truth?

Jack and Jill went up the hill to fetch a random sample,

A deer looked on, them goats, they yawned, & beetles they found ample.

You, I trust, are not a fool. I can assure you, I’m no experienced journalist, either. But for example’s sake, let’s just say that you are slightly unintelligent and I have an ulterior motive. I want you to believe every idea I put forth, in your favourite newspaper. I’ll make them believable, by doing what the devil does - refine the lie, with a grain of truth.  

I want you, and the other unsuspecting buffoons, to fall for my lie.

Don’t freak out, just yet. All I’m trying to do is, draw your attention to the fact that, much of what you read, is probably much of what’s not true. And unfortunately, statistics in the wrong man’s  hands, has played a big part in messing things up.  

Allow me to show you how.

Mr. Ethan Greenfield, is the proud owner of the new pop-up vegan restaurant, ‘The Green Plate’.  He decides to circulate a survey that assesses the reasons why his customers chose to adopt a plant-based diet. Out of the 1000 customers to whom the survey was mailed, only 300 responded. These 300 customers are likely the ones who are passionate about veganism, and have answered with reasons that support the morals behind the diet. Greenfield, taking this into account, plans to have his menu designed with meals that are ethically crafted. He could be prepping himself for some big losses.

What about the other 700 customers, who never responded?                                                                They might have been the ones who adopted veganism due to health concerns. Or it just might be the case that, a now, not-so-happy customer simply preferred crunchy bacon to bland lettuce.          

It never got recorded.

Let’s move on to how felines, can be terrible for samples.

Cats. They’re all around her. You desperately look for signs of another option to choose from, but she insists. She’s curious to see if you’re fond of cats. Helplessly, you choose to let her hear what she is hoping for (and not lose your chances with her, but that’s another story), ending up with something that sounds like this: ‘Oh, how I adore them!’                                                                           

Truth is, you got scarred by one, and wouldn’t mind hissing at the next one you see passing by.

You can imagine, how many samples could have been biased, just based off of the interviewer. There have been numerous instances, wherein, just for the sake of sounding good, one responds dishonestly, or with a certain amount of falsehood. The magazine you’ve subscribed to. Your financial income. The list goes on.

The infamous 2016 Presidential Elections, showcased one such scenario. Hilary Clinton was expected to defeat Donald Trump, by a landslide and that’s what the exit polls indicated. However, the non-response of the ‘Shy Trump voters’ was what amplified this result, that being a major reason why the polls were so inaccurate.

There exists a plethora of factors that can hinder a sample from being completely random, and it can skew the results to quite an extent. Which is why, the next time you read a news article with a lot of numbers, one must consciously ask the question,                                                                                'Which percentage of the population does this alarming statistic signify?’            

Ensuring you understand the numbers right, and whether they can even be labelled, ‘right’, will reflect a favourable shift in the way you perceive information.

Bias is hated by all. Especially statisticians. If you mentioned you liked dogs instead of cats, the statisticians have lined up to congratulate you.

The lions too, Neyman agreed, would be a startling add-in,

And down the hill, went Jack and Jill, with quite a crazy sample.