Posted on:

02 Oct 2024

0

Doubt about A/B segmentation

Good morning,

I have a question about how to perform segmentation in an A/B test to check if a new feature generates discrimination. I have researched, and one of the suggestions I have seen is to observe the interaction effect. To simplify, let's assume we want to study if the feature being tested has a different impact based on gender: 0 Male, 1 Female, and group 0 Control, 1 Test.

To do this, we would generate the regression model:
Y = B0 + B1Group + B2Gender + B3(Gender * Group)

It is this last variable that would capture whether the effect of being in the test or control group is different based on whether you are male or female.

However, I have a doubt about this example. In the variable (Gender * Group), there are 2 values:
1 --> Female - Test
0 --> Female - Control, Male - Control, Male - Test

Therefore, even if B3 is significant, it may be due to a significant difference only between Female-Test and Female-Control. This is why I have doubts about how the segmentation discussed in the video would be applied.

Best regards, and thank you in advance.
0 answers ( 0 marked as helpful)

Submit an answer