CLPEX Chat Board

...at a 94% correct identification rate, and a 3% false positive and 3% false negative rate. Based on that, they are actually better at exclusions that we are!

This was just published and the context is how well a dog can sniff Covid19 from a sample.

https://bmcinfectdis.biomedcentral.com/ ... 20-05281-3

Within randomised and automated 1012 sample presentations, dogs achieved an overall average detection rate of 94% (±3.4%) with 157 correct indications of positive, 792 correct rejections of negative, 33 false positive and 30 false negative indications.

This was interesting, however, when talking about acceptable errors in diagnostic testing:

...the current gold standard diagnostic RT-PCR test of a nasopharyngeal swab can, in trained hands, have a false detection rate of 25% and a false positive rate of 2.3–6.9%

So how do we match up with the medical field?

As I read the study, the 94% accuracy with 3% false positives and 3% false negatives was the average "hit rate" of single dogs. I wonder if you could improve that with blind verifications? What is the accuracy rate if a sample is run by two dogs working separately and they both report the same results? If the two dogs disagree, try a third for best two out of three? Seriously here, just wondering.

Let's see how they do with a 5-conclusion scale.

Apparently, the dogs are getting better. Some are now claiming a 100% success rate. Better look out, Dogs. Look at what happened to the fingerprint people when Stephen Meagher made that claim about fingerprints at the Daubert hearing in the Byron Mitchell case in 1999. The doggie OSAC will be all over you for making unscientific claims.

https://theconversation.com/these-dogs- ... ate-143756