I have been passed this from another poster on Amazon.
To put that into some kind of context, many clinical tests are nowhere near that reliable.
If the mods want to move it to another thread, that is fine.
----------------------------------------------------------------------------------------------
You might find this useful when discussing the dogs and the ''Incredibly unreliable'' claim, which was demonstrably false
http://www.caninesearchsolutions.org/wisconsin_v._zapata.pdfIn particular, this paragraph
''* The negative response to the question of the canines' detection reliability was based
upon the dogs' actual searches in the field. The judge concluded that any/each time a
cadaver dog searched and cleared an area without providing an indication and no person
or other tool returned to the search location to verify that there WASN'T something
there, the dog's conclusion that the area did not contain the odor of human remains had
not been confirmed and was therefore was not reliable. ''
This was the basic problem with quoting the Zapata case. The measure the judge used to determine reliability was neither scientifically nor statistically valid.
If the dog came back after a search having not registed a hit, there are only two possibilities
1) The dog is correct, there is no evidence of cadaver, ie a ''true negative'' or
2) The dog is incorrect, it has missed the evidence, and this was therefore a ''false negative''
However, in the Zapata case the judge decided that unless someone had actually checked the area again, each 'negative' was a 'false negative'
Now clearly, this is absolutely not a valid conclusion, but using this strange formula has the effect of reducing the apparent reliability of the dogs considerably. In effect it is saying ''the dog is only right if it finds something'', and takes no account of the searches when the dog doesn't find anything because there is nothing to find. It also raises the issue that the subsequent searches are equally only as good as the dog searching - unless he thought he could send out a handler with particuarly keen senses.
It also has little validity with respect to Zapata because it concerns purely 'false' negatives, and not false positives.Likewise, in the McCann case the issue concerns positive indications, which studies have shown as being of the order of 90% accurate.
In the Zapata case the dogs alerted to locations the accused subsequently confirmed he had moved the body to.
The data from the Oesterhelweg et al study is far more reliable as the study was in controlled conditions and compared to a non-cadaver. In summary, that showed the following:
Sensitivity (75-100), specificity (91-100), positive predictive value (90-100), negative predictive value (90-100), accuracy (92-100).
To put that into some kind of context, many clinical tests are nowhere near that reliable.