The Shape of the Data#
SLOP1638 · Week 9
You did not start this on a good day#
Nobody takes up a sleep routine during a week they slept beautifully.
Nobody reorganises their day on a day that went well.
One measure, thirty days#
Thirty days of one measure. Illustrative data, not a real record.The worst day, and the three after it#
The lowest day, and the three that followed it.Unusual measurements tend to be followed by ordinary ones#
So anything you begin at the bottom looks like it worked#
Not sometimes. Reliably, and convincingly, from the inside, because you genuinely did feel better afterwards.
You are not being fooled about how you felt.
You are being fooled about why.
The flight instructors#
Instructors noticed that praising a cadet after an exceptional landing made the next one worse. Criticising one after a bad landing made the next one better.
They built a training philosophy on it.
An exceptional landing is followed by a more ordinary one. So is a terrible one.
What this does to your own week 5 result#
If you began your experiment on an unusual day, part of your result is the day being unusual.
Go back to your week 5 record. Find your strangest day. Look at the day after it.
That is the size of the effect you would have claimed.
Two things that help#
A control condition. Something to compare against that also started at the bottom.
A pre-registration. Written before you began, so the question cannot move once you have seen the answer.
Neither removes the problem. Both make it visible.
In the lab today#
We pool the class’s week 5 records, find every unusually bad day, and look at what followed.
No statistics. Just the shape of the data.