In this episode, we discuss how dangerous statistics can be.  People often confuse statistics with facts, but anytime someone quotes a statistic, there are a few things you need to keep in mind.

This topic has bonus content

Thanks for joining me for another podcast.  I really appreciate all my new subscribersand those of you that have left reviews on iTunes and other places.  As always, I appreciate both positive and constructivesuggestions. 

Today’s episode is:

Statistics: 9 common ways the numbers are lying to you

I’m not going to lie, this has been the most difficultpodcast to produce so far.  I’m not astatistician, and trying to simplify complex statistical concepts is definitelytricky.  I have at least 15 hours intothe production of this episode.  I hopeyou enjoy it.

The content in this podcast is adapted from How to liewith statistics by Darrell Huff.

As I mentioned in my intro podcast, statistics can actuallybe quite dangerous

As a young grad student, someone much older and wiser thanme recommended I read How to lie with statistics.  It was spot on when it was first printed in1954 and is still very relevant today

Today I am going to discuss some common pitfalls we can runinto when interpreting statistics

Statistics pitfall #1: Sampling bias

For a study to be accurate, it must faithfully represent thepopulation of interest.  For example, apolitical telephone survey that uses only landline numbers misses a largeportion of the US population who only have cell phones.  No matter how much data a study collects, ifit doesn’t represent the group in question, it is pretty worthless.

The problem with studying humans, is that we can’t justrandomly assign people to different groups for the sake of science and soalmost all study participants self-select to some extent

This sampling bias can be conscious, or unconscious.  We can correct to some extent for consciousbias, but unconscious bias is particularly dangerous because we often don’trealize it is happening.

For example, when YouTube created a new video loadingfeature, about 10% of videos were loaded upside down.  When they began trouble shooting why so manyusers loaded them incorrectly, they discovered that most of the upside-downvideos belonged to left-handed people. Because of an unconscious bias towards right-handed people, the app leftout about 10% of the population. https://www.eliinc.com/five-real-world-examples-of-unconscious-bias/

Another example of self-selection bias occurred in BostonUniversity’s study of brain trauma in American football players.  The results of the study were widely reportedas “99% of football players had CTE” even though the researchers admitted thatthe study population was biased.  How wasit biased? All of the 202 brains examined in the study were from players whoexhibited neurological symptoms while living. For the results of the study to be accurate to the population, brainswould have to be taken across a wide range of people who had at one point intheir life played football, not just those with symptoms.  Additionally, nearly half of the brainsstudied were from professional football players, which is a very small subsetof those who play football. https://greatbrook.com/biased-survey-samples/

Statistics pitfall #2: Different types of averages

Another common pitfall is using the wrong type of average

In statistics there are 3 types of averages: mean, medianand mode

Mean is the classic average that we are all used to: add upall the numbers and then divide by total numbers used

Median is the middle value of a group of responses (fiveresponses, order high-low, it is response #3)

Mode is essentially the most common number in a group ofresponses

Why does this matter you might ask?

If you look at a graph of how long Australians live, themajority of people live longer than average. How does that work?  This is