confidence values
Message posted by lee (via 141.163.167.190) on December 20, 2001 at 11:43 AM (ET)

Hi

I hope someone can help with this cos its driving us all insane :)

Weve got over 300 images of similar looking fish. Weve shown these to various experts and theyve classified them accordingly. Naturally some of the images have 100% agreement amongst everyone while others are some way off that. We dont have a baseline because the only way to get one would be to trust entirely the judgement of one expert over another.

What I'd like to find out is if there is a way of getting more from this data other than saying the classifications are better than chance? Most of the images had an agreement of over 80% so are quite high - but these are experts so thats what we expected. Is there some more stringent testing or another way we can get a cofidence level from the data that we can be more confident in?

thanx

lee

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: confidence values
Message posted by JG (via 128.8.23.221) on December 20, 2001 at 5:31 PM (ET)