When the lone 80-year old that participated is weighted against the hundreds 20-year olds (I'm exaggerating, I hope) we still don't know whether her opinion is representative for the other 80-year olds that didn't respond. There is no point of weighing her more than a 20-year old.

My glib comment was about the fact that before weighing, I could have at least attempted to imagine who would typically respond to such an online-survey. It's a biased sample. But it's a sample. I could say: Well at least this many people out of the whole population answered this or that way. Once weighing mixes things up I don't even know how people responded anymore.