Free explainer2 min read
Types of bias
Understand the model, recognise it in a stem, separate the look-alikes, then apply it.
Watch it explainedThe study had ten thousand participants. It is still wrong, and ten thousand more would not help.
01Memory anchor
A marathon results board, rigged from three separate angles.
1. THE GATE MARSHAL ONLY LETS FIT-LOOKING RUNNERS THROUGH THE STARTING CORRAL, TURNING AWAY ANYONE LIMPING -> selection biasA systematic error from how participants were chosen, making the study sample unrepresentative.Read more →: a systematic difference in who is sampled or retained.
2. THE FINISH-LINE STOPWATCH IS STUCK, ADDING THE EXACT SAME TEN SECONDS TO EVERY RUNNER'S TIME -> biasA systematic error that skews a study's results away from the truth in a particular direction.Read more → is a systematic, direction-consistent distortion, not random scatter.
3. TIMING TEN TIMES AS MANY RUNNERS ON THAT SAME STUCK STOPWATCH JUST MAKES THE WRONG AVERAGE MORE CONFIDENT -> more data does not fix bias; only fixing the marshal's gate or the stopwatch itself does.
4. A RAINSTORM THAT DAY GENUINELY SLOWS EVERY RUNNER DOWN -> confoundingWhen a third, unaccounted-for factor influences both the exposure and the outcome, distorting the apparent link between them.Read more →: a real third factor affecting the outcome, not a bias to correct.