when you randomize into treatment + control groups, the groups are the same IN EXPECTATION, not IN OBSERVATION.

it's possible your groups will have different observed characteristics, it doesn't (always) mean "randomization didn't work"

Comments