Conversation

To me, it's obvious that limited data is better than no data, and you interpret it in context, evaluate theories for why you got that data (including stuff like sampling bias, etc.) But I'm starting to suspect most people view data as being a binary support or not of a view?

323

Aella

@Aella_Girl

Dec 5, 2022

like, to me I can publish a little bit of data and I just default assume everyone knows that I'm not drawing hard conclusions from it, it's just a piece of evidence we can discuss, but people often respond like I think I'm making a bold, irrefutable claim about reality

184

Aella

@Aella_Girl

Dec 5, 2022

which... feels a little startling to me. Should I start placing "I'm not definitively assuming anything from this data, a little data is better than no data" disclaimer at the beginning of everything? I just had assumed sharing the sample size and source was serving that function

154

Replying to

bad data is often worse than no data

Aella

@Aella_Girl

Replying to

@SouthSidelarry

there is no bad data, as long as its not like, totally fradulent or generated via a glitch or whatever. The badness is in the interpretation.

8:16 PM · Dec 5, 2022

Replying to

And the predicate for that interpretation is the bad data.

Replying to

and

Pretty sure an open Google survey with zero ID verification or controls is the exact definition of bad data but sure

Replying to

and

I agree, but I wonder how likely it is that wrong conclusions will be drawn from weak data? If you can predict with high likelihood that most people will misunderstand and/or misuse the data, what is the correct move to make?