Conversation

Also curious what sample size u guys would intuitively guess is good, given the questions in this. I'm aiming for around 5k, but I'm suspecting I might need more than that in order to get enough data from e.g. ppl with longer relationships

Quote Tweet

Aella

@Aella_Girl

Dec 24, 2022

If you haven't yet taken my new relationships survey, I'd love if you could! Answer questions about your romantic relationship and see what fictional couple you're closest to. guidedtrack.com/programs/lo4g4

Replying to

the size isn’t the issue, the issue is going to be representativeness. it will be representative of people who see and take surveys you make, which is only a representative sample of precisely that group

Replying to

You don't need a perfectly random sample for useful data, jfc

and other critiques

1:00 AM · Dec 25, 2022

293

Views

Replying to

this is wrong, sorry. or rather it’s a straw man: the issue isn’t that you need total randomness, it’s that the sample you’re going to get is going to be wildly self-selecting and I doubt you’re even doing the limited adjustments you could do with that kind of data set

Replying to

Data this imperfect is hardly any better than no data at all. Try to set some reasonable priors on results from subgroups and see where it gets you. If data aligns with priors, you don’t need data, and if it doesn’t, sample non-representativeness will not let you move forward.