Subgroup findings are like noticing a particularly bright thread in a sample swatch—it suggests a potential pattern, but you need to weave a whole new length of cloth to confirm it’s not a fluke. The ISIS-2 star-sign finding perfectly showed that; a random subgroup can show ‘effect’ purely by chance. In my work, I’ve seen a drug seem to work wonders in, say, left-handed patients in a small trial, only to vanish when we test it properly. So now, we never let a subgroup hypothesis become a result until it’s been deliberately put on the loom for a new trial.
#random#life
Jump in to reply — no account needed.