say you want to try adding two features, which don't you think interact with each other, e.g. a change to the "pick this room button" and a change to the checkout flow. then, you can randomly assign users to two experiments, independently. your t-test results should then be valid if the two features are independent.
your t-test results should then be valid if the two features are independent.
Assuming that your assumption are correct on interaction effect.
You can do a hypothesis test on that assumption while including both factors (the two features). Which will clear away any doubt with a 95% confidence or hire a statistician =).
Comments
say you want to try adding two features, which don't you think interact with each other, e.g. a change to the "pick this room button" and a change to the checkout flow. then, you can randomly assign users to two experiments, independently. your t-test results should then be valid if the two features are independent.
Assuming that your assumption are correct on interaction effect.
You can do a hypothesis test on that assumption while including both factors (the two features). Which will clear away any doubt with a 95% confidence or hire a statistician =).