Here’s the tricky one because while the others are similar to what by brain imagines, I just can’t visualize clearly how adjusting for a collider plays out
I find that this example helps too. If the true relationship between niceness and appearance is zero, but you run a regression by only looking at people you've dated (i.e. adjusted for that collider), you're missing a whole quadrant and the result is wrong and only valid for part of the population
Comments