Monday, October 31, 2011

To “Nanny-Dog” or “Not To Nanny-Dog,” that is not the question

To “Nanny-Dog” or “Not To Nanny-Dog,” that is not the question

To continue the talk of the deaf....

Science, or scientific research, does not answer the question ‘why.’ Scientific inquiry deals with probabilities. For example, if event “A” happens, what is the likelihood that it will be followed by event “B” or event “X?”

So, if we call the Pit Bull a “Nanny Dog” what’s the probability of someone getting bitten or killed by a Pit Bull? Conversely, if we agree that Pit Bulls are too dangerous to have ever been called “Nanny Dogs,” what is the probability that someone will show up at my door and want to take my dogs away and kill them?

So let me set up a decision making matrix to see which way I would rather err..

The decision-making matrix will depend on the payoff.

Null Hypothesis (H0) is true
Nanny Dogs
Alternative Hypothesis (H1) is true
Not Nanny Dogs
Accept Null Hypothesis
Nanny Dogs
Right decision
Wrong decision
Type II Error
Reject Null Hypothesis
Not Nanny Dogs
Wrong decision
Type I Error
Right decision

H0 is False and I call it False – we continue calling Pit Bulls “Nanny Dogs.” All Pit Bulls are loved and owned by responsible families. If we make a Type II error – some people get bitten, maimed, or killed. The proposition is to mitigate the Type II error by educating the public and make them “bite-proof.” Which is what the experts tell us to do. The number of Pit Bull (or any dog) bites decreases.

Ho is true and we fail to reject it – we state that Pit Bulls are not “Nanny Dogs” (because… they are vicious – otherwise why not call them potential “Nannies?”). Millions of dogs get killed, ownership is penalized, more governmental control is put in place, people learn to be helpless because somebody else will take care of them, individual responsibility is, once again, minimized, and OTHER type of dogs will bite, maim and kill. A Type I error is vastly most costly, if you care about dogs, freedom, responsibility, and are one of the hundreds of thousand of us who love Pit Bulls.

My opponents’ aim is “prove” that the Pit Bull was not named a “Nanny Dog” (because “it couldn’t possibly, look at the terrible evidence we have..”) thereby further demonizing an animal that doesn’t deserve it. My projection of his/her desired outcome: nothing will change, except hundreds of thousands of more dogs will be killed. Cities that banned Pit Bulls have NOT seen a decrease in dog bite fatalities. I can bring statistics supportive of this statement. Their supporters: people who have been bitten by dogs, people who are afraid of their shadows and want Big Brother to protect them, politicians, media, and breeders.

My aim is educate the public and let it know that the dogs are safe when handled properly and the public needs to learn some basic knowledge to protect itself against DOG bites. I said, DOG, not Pit Bull, on purpose.

The focus of Bully-haters remains heavily on the dog’s conduct rather than on the owner’s conduct, which, to me, seems misguided. Owner conduct is easier to correct through law, education or other means, which is more likely to promote owner accountability for dogs in the future. Focusing on the dog’s actions may mean it is destroyed as ‘dangerous’ while the owner can still get a new dog, a new kind of dog, or an alligator and act equally irresponsibly in the future.

Calling Pit Bulls “Nanny Dogs” seems to irk a cadre of Bully-haters to such extreme that makes me wonder not only about their agenda, but also about their mental health. In a country in which up to 300 kids are killed each year by their biological parents we worry about what a Pit Bull may or not have been called? What a costly investment of misplaced priorities, energy, time and resources this is.

Just in case I have not stated my intentions clearly, the goals of this blog have always been and remain to:

1.      Debunk the bad rap my favorite dog breed gets;
2.      Provide (however biased) evidence that Pit Bulls have been vilified by the media ad nauseam;
3.      Present evidence that close-minded  people use b.s. published in the media to further their own agenda with no regard to whom it may harm;
4.      Provide whatever means I can muster to show off Pit Bulls as the great pet that they are in the right hands and with proper treatment;
5.      Remind the uneducated that the breed was bred for dog-aggression NOT human-aggression and that two are vastly different from each other;
6.      Help the Pit Bull by convincing folks to a) spay or neuter their dogs; b) understand and assume responsibility of dog ownership; c) behave as they want others behave towards them; d) put unethical and immoral breeders out of business; e) stop dog-fighting; f) prevent the killing of nearly 500,000 Pit Bulls every year by rescuing, fostering, training and adopting as many of them as possible; and g) stop ill-equipped, uneducated, mean-spirited, angry individuals from showing up at MY doorstep and take and kill my dogs simply because they look like something a politician thought of as not deserving to live

So, it’s not about “Nanny Dogs” at all. It’s about BSL, stupid!

More about that later.


DubV said...

Part 1

Well, it isn't everyday that I have a blog post somewhat written to me, so forgive the long response.

What you mean precisely by your nanny dog claim is becoming less clear. Therefore, your hypotheses seem underspecified. I'll adopt your classical frequentist statistical approach here, and I'll also hazard a guess that your precise hypothesis is "pit bulls were commonly called nanny dogs during period X". Period X is the bygone era you reference. If we are to keep your methodology of imposing a sharp dichotomy on reality (which your column structure surely implies regardless of whether or not matrix cells are seen as having probabilities assigned other than 0 or 1), then we would be forced to adopt a non-directional null hypothesis. That null would be "pit bulls were NOT commonly called nanny dogs during period X". Notice that this null does not correspond to your description of the null hypothesis you incorporate as a possibility. In fact, your null hypothesis is not a null hypothesis at all. It is either another alternate hypothesis or else it could be cast as a special case of the null hypothesis I have given in quotation marks in this paragraph.

Your two column layout with descriptions of both the null and alternate hypotheses is therefore an obvious false dichotomy but the column structure and description does not yet qualify as a straw manning attempt (that comes later when you state that I am attempting to prove the null hypothesis you offer). If we wanted to be more comprehensive and include your attempt at a "null" hypothesis we would need a minimum of three hypotheses and three corresponding columns: 1. during period X, pit bulls were commonly called nanny dogs, 2. during period X, pit bulls were NOT commonly called nanny dogs for reasons unrelated to temperament (e.g. no one thought to call them that, they were one among many dogs good with kids, or other), and 3. during period X, pit bulls were NOT called nanny dogs for reasons related to their poor temperament in this regard.

DubV said...

Part 2

Now for the rows. The rows in your matrix correspond to various beliefs or perhaps thought actions that may lead to physical action. If we keep two columns and avoid a false dichotomy, then the beliefs or thought actions revolve around the null (PB were NOT commonly called nanny dogs in period X) and we will need two rows as well. The terminology of null hypotheses and the like are often misunderstood by the general public and scientists alike, so I will expand a bit. Because the null is cast as a negative truth claim, we never accept the null hypothesis because a universal, negative truth claim (such as the null is purposefully cast) cannot, by definition, be proved only disproved (you reflect this in the terminology used in your matrix when you use the terms “reject the null” and “fail to reject the null”). If there is sufficient evidence for the alternate hypothesis (PB were called nanny dogs during period X) then the act of rejecting the null moves us by default to this positive truth claim in some sense (I am not the first to notice that the reasoning behind frequentist statistics is convoluted, but it is how you cast this blog post). If there is not sufficient evidence to reject the null then we stay in our holding position of not rejecting the null. My position regarding evidence of your positive truth claim (PB were called nanny dogs in period X) is therefore only a positive truth claim regarding your evidence, but in regards to reality it is a position of indifference as an object still under its own inertia. My position is consistent with not rejecting the null as it is purposefully in these situations stated as a position that cannot be proved, only disproved with sufficient evidence. I mention this not to give you an elementary lesson in statistics, but to combat your false assertion that I am attempting to prove the null hypothesis in this case. The entire basis of frequentists statistics is tied to the alternate hypothesis (what you claim) as being the provable assertion and so holding the burden of proof (this is also apparent in the mathematics related to constructing confidence intervals).

DubV said...

Part 3

I mentioned before that to keep your attempt at a "null" hypothesis we would need to expand your matrix to use three columns in order to avoid a false dichotomy. If we took this tack, then we could also employ three rows as well, each representing the position we feel is best supported by the evidence. This would move us out of the realm of a simple frequentist approach, but we can explore this as a thought experiment. The rows now could not be simply represented as rejecting or failing to reject the null. In this case, there are two unprovable assertions and your claim still holds the burden of proof. This is because if I cannot prove the negative claim that "PB were NOT called nanny dogs in period X" (which I hope is now clear) then I cannot possibly both prove this negative and give a reason (such as a poor temperament). Therefore, you are left where you started, holding the burden of proof which is plainly obvious. Support for your claim "in period X, PB were called nanny dogs", if all three are on the table, would cause us to reject the other two alternatives. So you are still left holding an empty sack.

DubV said...

Part 4

Regarding the later part where you evaluate your decision matrix, I am loath to approach this one as it is an emotional bit and reaches its tentacles too far beyond the original discussion. Also, I will be forced to enter into how you have framed things somewhat, which you may incorrectly take as me agreeing somewhat. Also, it somewhat casts decisions related to reality as being a group endeavor which is incorrect, but I suppose will do when deciding what beliefs to encourage in others. Nonetheless, here goes.

In decision theory, you first estimate the probabilities of possible outcomes related to your potential decisions and then intersect this with a utility value which can be applied by the analyst, a group of stakeholders, etc. The aim is to select the decision with the highest expected net benefit or utility. What is suspiciously absent from your utility calculations is the value of actually being correct! It seems you take a consequentialist, normative view of truth, which I strongly disagree with for many reasons (ever hear of Leo Strauss?). As far as what you expect to happen if various squares in the matrix are occupied, it seems you hang a huge and heavy weight on this nanny dog claim. Do you honestly believe that people believing pit bulls were once called nanny dogs is that important? I simply cannot see so much riding upon this one truth claim.

DubV said...

Part 5

In regards to the specifics, you should edit your blog. Your description of a "false alarm" and "miss" are backward. In your conception as otherwise framed, a "miss" would be pit bulls actually being nanny dogs in the past and NOT calling them that (which is not what you write and "false alarm" is wrong for a reason you should see now). About what you call a "miss" here (I'll keep your verbiage and not speak of it correctly as that would add another layer of confusion) which you say results from people incorrectly thinking pit bulls are nanny dogs when they are not; the backbone of your matrix argument here as given is what would be the right course of action without reference to the probability of us occupying a given matrix cell given our decision. Am I correct to state that even if the data showed this "miss" to be somewhat likely, that you would go on having people believe pit bulls are nanny dogs but rely upon education to make the public "bite-proof"? If that is true, it is hideously immoral. This is especially true given that how you actually misrepresent the null hypothesis here (as pit bulls being vicious in part) and that the nanny dog claim specifically addresses how well the dog does with children who are simultaneously: 1. least able to defend themselves, 2. comparatively innocent, and 3. less likely to be involved in dog breed choice, whether to get a dog, or how the dog is kept.

Now on to what you call a "false alarm" which you describe as pit bulls being nanny dogs and the public not believing this. This goes back to what I said before about my doubts regarding the huge importance of this nanny dog claim. I simply do not think reality would change much if this particular meme was subtracted from our culture. Do you honestly believe that pit bull adoption rates would plummet without this and that pet owners would be affected to a huge degree? I have no absolute proof this is not the case, but it just seems wrong on its face. Have pit bull euthanasia rates appreciably dropped since the nanny dog claim has become more widespread?

DubV said...

Part 6

In frequentist statistics, selecting a cut-off level for deciding between your alternative beliefs or actions is intimately tied to balancing the probability of type 1 and type 2 errors (your misses and false alarms). When you intersect these error rates with utility, you are at least implicitly weighting the importance of each error type. The way you have described it, an error of one type predominantly affects (likely innocent) humans while the other predominantly affects pit bulls. Perhaps many more pit bulls would be affected than humans, but still your calculus must involve a weighting of human versus pit bull interests if your two error types are counterposed as you make clear. So, I must ask you: how many pit bulls would you euthanize if guaranteed to prevent the mauling of one child? I have an answer. I would euthanize every pit bull that has ever been saved by your rescue to prevent one child’s life from being permanently changed by a mauling, but that's just me.

Now about those "hits" and "correct outcomes". Again, I value truth for its own sake. Your correct rejection scenario implies that pit bulls currently are nanny dogs. You state if this was believed by all then "all Pit Bulls are loved and owned by responsible families". You cannot possibly believe that correctly believing a true nanny dog claim is either a sufficient or necessary cause to bring what I have quoted into existence can you? As far as you not believing the nanny dog claim and it not being true, remember how you personally misrepresented the null hypothesis (pit bulls are vicious). Next, do you honestly believe some 1984-esque, slippery slope phenomena would unfold? If the null was true as you misrepresent it and this means in your decision matrix cell lie vicious animals, do you believe it would be wrong to have higher euthanasia rates in that case for those animals?

DubV said...

That may be an affront, but it is certainly not an ad hominem. An ad hominem is not an insult, it is a very specific form of fallacious reasoning. I do not apologize for being blunt. A trained scientist could only make the errors you did on purpose. And suppressing a sharp critique which exposes your errors speaks for itself.

DubV said...

Don't forget Part 7

What I have said in the last few paragraphs dips into your framing of the null hypothesis, which I have already shown to be faulty. Failing to reject the legitimate null hypothesis, and having people go by this, simply means people stop repeating the nanny dog claim because it is not supported by the evidence. If this decision aligns with reality, then we all have a bit more truth in our lives and perhaps fewer parents are lulled into a false security of owning a large, fighting breed and leaving it with their kids. If this decision is incorrect, one thing deprived is one of several arguments that pit bull advocates use for breed image rehabilitation. This would be incorrect by omission and would not be ideal so far as it is incorrect. Also, perhaps this would marginally lower adoption rates. If we reject the null, then pit bull advocates will go on with the nanny dog claim. If they are correct in this, then good I would be happy to realize I was wrong and for greater truth to be present in the world and maybe a few deserving nanny dogs are helped. If we reject the null and are wrong, then pit bull advocates spread an untruth which may end up endangering a certain number of people, not to mention other animals that might be attacked.

As far as your other claims, including that banning pit bulls has not decreased dog bite related fatalities, those are unrelated to this discussion. Therefore, I won’t reply to that at this time.

P.S. Frequentist stats, which you use here by invoking rejection of the null, relies upon crisp decision making regarding mechanism and hence assumes causality. This is because we are only looking at the probability of the data GIVEN the model, unlike Bayesian stats where one is able to infer the probability of the model GIVEN the data.