Re:Court rules drug dog barely more accurate than a coin fli...
by shkkmo on Friday August 07, 2015 @01:47PM
(#10023831)
This headline is pretty terrible.
Keep in mind when looking at these statics that the number for this dog is not based on a random sample. Therefor it is impossible to actually gage the dog's accuracy from these numbers, since we don't know the ration of false:true negatives, only the number of false:true positives. The ratios could be identical, which would make the dog's alerting behavior random. There could even be a higher ratio of false:true negatives, which would make the dog worse than random.
The only meaningful number is the 93% ratio which basically means that the dog is very close to being a 'probably cause generator' as mentioned in the article.
Keep in mind when looking at these statics that the number for this dog is not based on a random sample. Therefor it is impossible to actually gage the dog's accuracy from these numbers, since we don't know the ration of false:true negatives, only the number of false:true positives. The ratios could be identical, which would make the dog's alerting behavior random. There could even be a higher ratio of false:true negatives, which would make the dog worse than random.
The only meaningful number is the 93% ratio which basically means that the dog is very close to being a 'probably cause generator' as mentioned in the article.