CSI and cadaver dogs - some facts and statistics.

[OxfordBloo]

I understand squaring numbers I did that for 11+. and I understand basic probability all though it is unclear why you consider the parallel events have a dependent relationship.
You say however "most papers suggest" but you are unable or unwilling to show where or what those papers are. Saying Google is your friend don't cut it. You could be kind enough to tell us which papers your opinion is reliant on. It is a simple enough request.

Anyone suggesting that the dogs are perfect needs to produce a blinded peer reviewed paper.

There is no such paper.

As all papers show an uncertainty of between 60% and 90% , no matter what the absolute value is, it increases with the parallel status of the trial.

[slartibartfast]

My take for your information.

[G-Unit]

No. My contention simply is that a joint Eddie Keela alert has a higher uncertainty than a single alert whatever the actual vale of certainty for a single alert.

80% is a fair average of the published papers, but that is not my point.

The point is that the uncertainty increases because of the parallel nature of the test.

In this case  Keela's score was 100% correct.

[Alice Purjorick]

Anyone suggesting that the dogs are perfect needs to produce a blinded peer reviewed paper.

There is no such paper.

As all papers show an uncertainty of between 60% and 90% , no matter what the absolute value is, it increases with the parallel status of the trial.

I did not say that.
Several posts ago you maintained you had read all there was worth reading about cadaver dogs and indeed had communicated with one of the worlds leading experts. It should not be too difficult to find a reference in all that lot and tell us should it?

[G-Unit]

No. My contention simply is that a joint Eddie Keela alert has a higher uncertainty than a single alert whatever the actual vale of certainty for a single alert.

80% is a fair average of the published papers, but that is not my point.

The point is that the uncertainty increases because of the parallel nature of the test.

That is the assumption which requires verification, because that's the assumption upon which your conclusions rely.

[OxfordBloo]

That is the assumption which requires verification, because that's the assumption upon which your conclusions rely.

No. I am not assuming that dogs are 80% accurate. Dogs are either 100%accurate or they are less than 100% accurate.

No evidence exists for them being 100% accurate. Blinded peer reviewed pets suggest 60% to 90% accurate.

My contention is not that we can prove that dogs are x per cent accurate, but that two dog trials for blood and or Cadaver alert reduces whatever the accuracy is by the mutpilicand of the two uncertainties, thus reducing the reliability.

So we ca say that IF the single probability is 'x' then the twin probability is x squared, and as x is always less than one, the double alert is less reliable than any single alert at that probability.

I am not claiming that dogs are 80% reliable. I am saying IF the probability of each single alert is 80%, then the probability of the double alert is 64%. If it is 95% it is 90% if it is 70% it is 50% and so on.

It just so happens that 80% is a fair average from all blinded peer reviewed papers on the subject.

[G-Unit]

No. I am not assuming that dogs are 80% accurate. Dogs are either 100%accurate or they are less than 100% accurate.

No evidence exists for them being 100% accurate. Blinded peer reviewed pets suggest 60% to 90% accurate.

My contention is not that we can prove that dogs are x per cent accurate, but that two dog trials for blood and or Cadaver alert reduces whatever the accuracy is by the mutpilicand of the two uncertainties, thus reducing the reliability.

So we ca say that IF the single probability is 'x' then the twin probability is x squared, and as x is always less than one, the double alert is less reliable than any single alert at that probability.

I am not claiming that dogs are 80% reliable. I am saying IF the probability of each single alert is 80%, then the probability of the double alert is 64%. If it is 95% it is 90% if it is 70% it is 50% and so on.

It just so happens that 80% is a fair average from all blinded peer reviewed papers on the subject.

Unfortunately, for it to have any meaning in this case you would have to demonstrate that Eddie and Keela in particular were not 100% accurate. Dogs in general may have been shown to be inaccurate in tests, but even that doesn't translate to being inaccurate 'in the field' because testing handlers and dogs has it's own problems.

[OxfordBloo]

Unfortunately, for it to have any meaning in this case you would have to demonstrate that Eddie and Keela in particular were not 100% accurate. Dogs in general may have been shown to be inaccurate in tests, but even that doesn't translate to being inaccurate 'in the field' because testing handlers and dogs has it's own problems.

If you think dogs are generally accurate, produce the evidence. All measurements whether doggie or not have an uncertainty.

I am not even talking about dogs.

I am talking about any parallel trial (see the solid hollow red blue spheres exile above). All parallel measurements result in reduction of accuracy. That is all I am saying.

[G-Unit]

If you think dogs are generally accurate, produce the evidence. All measurements whether doggie or not have an uncertainty.

I am not even talking about dogs.

I am talking about any parallel trial (see the solid hollow red blue spheres exile above). All parallel measurements result in reduction of accuracy. That is all I am saying.

I have asked you to produce your evidence for the general inaccuracy of the dogs and you haven't. It's your thesis not mine, so the onus to produce evidence to support it is on you.

[OxfordBloo]

I have asked you to produce your evidence for the general inaccuracy of the dogs and you haven't. It's your thesis not mine, so the onus to produce evidence to support it is on you.

My contention does not depend on any particular inaccuracy. It applies for any value from 99% to .1%.

[Carana]

In this case  Keela's score was 100% correct.

As there was no forensic corroboration of the substance, I don't see how it can be known whether Keela was correct or not.

[OxfordBloo]

As there was no forensic corroboration of the substance, I don't see how it can be known whether Keela was correct or not.

Correct.

[slartibartfast]

No. I am not assuming that dogs are 80% accurate. Dogs are either 100%accurate or they are less than 100% accurate.

No evidence exists for them being 100% accurate. Blinded peer reviewed pets suggest 60% to 90% accurate.

My contention is not that we can prove that dogs are x per cent accurate, but that two dog trials for blood and or Cadaver alert reduces whatever the accuracy is by the mutpilicand of the two uncertainties, thus reducing the reliability.

So we ca say that IF the single probability is 'x' then the twin probability is x squared, and as x is always less than one, the double alert is less reliable than any single alert at that probability.

I am not claiming that dogs are 80% reliable. I am saying IF the probability of each single alert is 80%, then the probability of the double alert is 64%. If it is 95% it is 90% if it is 70% it is 50% and so on.

It just so happens that 80% is a fair average from all blinded peer reviewed papers on the subject.

So no comment on my figures then?

[OxfordBloo]

So no comment on my figures then?

Ignoring the mathematics ands statistics ten.

[Carana]

So no comment on my figures then?

I don't understand how you came up with those percentages. Could you walk me through one or two examples so that I can get the idea?

One point that I believe skews the discussion is the assumption that "cadaver" scent necessarily means a deceased human (or pig).

Something I'm still not sure about is whether "human decomposition" scent is a euphemism or whether it is a more accurate term depending on the dog's training. I suspect the latter as we already know that Eddie reacted to dried blood from a living human being.

If a different dog had been solely trained on post-mortem human remains and had never reacted to decomposing material from a living human, then I'd find that issue clearer.

In other words, whatever the argument about percentages of accuracy, if a dog reacts to fruit, in the absence of corroborative information, you still don't know whether he's correctly alerted to an orange or a pear.

My conclusion being that Keela only reacts to bananas (and she may or may not always be accurate). Eddie reacted to fruit and if he reacted, Keela was wheeled in to see if she could detect a banana. Where Keela does not identify a banana, and Eddie reacts to fruit, how does anyone know which type in the absence of identifiable fruit pips or kernels?