I would like to start with terminoligy (so we all are on same page)
In the case of cadaver dog assuming it is a passive responce (I know yours is Nancy). I record correct responces 1. indicating by sitting/downing at target odor as a truepositive. 2. sniffing but not indicating on a controll true negative I record incorrect responces 3. Indicating on a control false positive 4. sniffing but not indicating on a target odor false negative 5. incomplete responces either as a falsepositive or a false negative depending on whether on target odor or a control. such as a major hesitation or momentay partial sit/down (this can be gray). The calculation would depend on how you train or test in my case I use controls, (duplicate containers and rewards) so calculation are made on numer of correct responces on both control and target odor are used to calculate, others may not do to method. But I thought I would throw out base terminoligy to see what others might use.
1. plus sign + for a positive, unassisted response.
To be a positive, unassisted response, the handler can not know the location, or actually, if anything is even there.
2. plus sign with a circle around it (+) indicates an "assisted response" .
Assisted response indicates, the dog is obviously in odor, but either can not determine source or will not give the final response. The handler is instructed as to the exact location of the target and told to get the cue response. Counts as a miss when figuring reliability.
3. Minus sign - indicates a miss. A miss is scored if the search is completed if the search area would have been vacated without anything being found.
Reliability is easy to figure. The number of positive responses divided by the number of targets placed.
Actual searches for drug detectors is figured in a similar fashion, number of "finds" divided by the number of responses.
False responses are determined by the number of false responses divided by the number of opportunities the dog had to respond correctly.
It's really a lot easier than it seems. It's a did or didn't scenario. I've found it very useful in court. There is no appearance of smoke and mirrors or some language that only master dog trainers can understand.
<<<< I was trying to figure some way that false repsonses would play into the reliability equation, but it looks like you treat them independantly.
False responses are part of the equation, yes they are reported independantly. For example; During a reporting period the dog the dog had the opportunity to find 20 targets. Let's say he found 19 and had one false response. To figure the reliability for that period divide number found by the number planted or 19 divided by 20. To calculate the false response you divide the number of false responses by the number of opportunities the dog had to make a correct response or 1 divided by 20.l
Our documented certification rate is NLT 90% proficiency with a less than 10% false response rate. We do a quarterly review of the records to determine the proficiency and false response rates.
<<<<Now do you start keeping relibaility AFTER the dog has attained some level of mastry (e.g., a certification) or right away?>>>
Reliability, or as we refer to as "proficiency rate" is tracke only after a dog has reached the established criterion for a particular odor. During intial training, each odor is taught seperately. The next odor will not be introduced until the dog has met the established criterion of; 10 consecutive, positive unassisted responses. The dog has then demonstrated learning that particular odor. Then the proficiency is measured. Once the dog learns all the odors to the criterion level, then it is certified. Once certified the proficiency level is documented on a quarterly basis. All dogs undergo an annual practical evaluation. Any dog falling below the documented proficiency rate during a quarter is removed from service, entered into a remedial training program, fixed or replaced.
I posted this last night but somehow that post got lost in the electronic nether regions of the internet.