AB testing with seasonability

Tomek

New Member
#1
Hello everyone!
I have the following task:
There are ~10k sportsmen trying to compete in a 100m running.

Some of them changed the diet on 15th July and we want to get to know if it affected their performance.
We want to know if that change of diet can make them faster.


But it's a bit tricky.
Their speed is seasonable. After 15 July it's getting hotter and hotter so usually, it takes more time for people to run under such conditions.

With the exception of sportsmen from Africa. For them it takes even less time when it's hotter outside :)


1564775039800.png


The question is: how to check if change of diet was beneficial for sportsmen?

I know that for data which is not seasonable, I could use Paired T test. But what test to do in that case?
Thank you in advance for your help :)
 

obh

Active Member
#2
Hi Tomek,

To me, it looks that you try to make it too simple ...

Why not adding temperature as an independent variable instead of the binary variable before/after 15 July?
Is the time after a change in the diet identical to all cases? will the change on diet influence the same a day after starting or 3 months after?