You want to predict the election. You call 1,000 households on landline phones
at 2:00 PM on a Tuesday. Who answers?
๐ญ How to Think About This
A "Sample" is a small slice of the pie. If you only slice the crust, you think the whole pie is
dry. You need to slice through everything to get the truth.
Who is in your sample?
Think about who has a landline these days. Think about who is
home at 2 PM on a Tuesday. Most people are at work or school.
Naive.
Your sample is heavily filtered. It is not random at all.
Workers are at work! They are not answering landlines in their living room. You
are missing 80% of the workforce.
Missing.
You missed the very people you were looking for.
In 1948, a poll predicted Dewey would win because they called people on
telephones. But in 1948, only rich people had telephones! The poll was biased
towards the rich. You made the same mistake.
Correct. Sampling Bias.
The Insight: It doesn't matter how big your sample is (1,000
or 1 million). If the method of collection selects a specific type of
person, your data is garbage. "Garbage In, Garbage Out."
The Strategy: "Who did you ask?" and "Who didn't you ask?"
๐ฐ The Slice
The Assumption
Error: Assumes access = universal.
The Reality
Fact: Access implies demographics.
๐ค Which thinking lens(es) did you use?
Select all the lenses you used:
๐จโ๐ฉโ๐ง For Parents & Teachers
๐ฑ Everyday Science
Kid: "Everyone loves this video game! All my friends play it."
Parent: "Who are your friends? Kids your age who like games?"
Kid: "Yeah."
Parent: "So 'Everyone' means 'Boys aged 10'. That's a biased sample. Ask Grandma if she
plays it."
LLOS.ai uses minimal cookies to remember your preferences and improve the site. Cookie policy.
LLLOS.AI
Welcome back
Choose how you’d like to sign in. Most people use Google — it’s the fastest.