Junk Science and Wiener Dogs

A "study" from the Center for the Interaction of Animals and Society at the University of Pennsylvania concludes that the "most aggressive" dogs are small breeds, and the most aggressive dog of all is the Dachshund.

The Dachshund? Wiener dogs are the most aggressive dogs on the planet?

Riiiiiiiight.

According to the authors of the study, after those killer Dachshunds comes the killer Chihuahuas, the killer Jack Russell terriers, and the killer Beagles.

Killer Beagles? Hmmmmmm. Something is seriously wrong here!

In terms of larger breeds, the highest aggression rating goes to the Akita. The Pit Bull came in sixth, while such notorious biting breeds as Dalmatians and Corgis did not even make it on to the list.

Whiskey- Tango- Foxtrot.

When you tell me Beagles are a dog to fear, and yet Corgis and Dalmatians have not even made it on to your breed survey, you have a credibility problem.

You also have a credibility problem when you list Huskies as one of the least aggressive dogs when, in fact, the Center for Disease Control counts them as among the top 5 dogs most likely to kill you (PDF).

The credibility problem is compounded when it turns out that this "study" is not based on post-dog-bite emergency room visits, but on something a good deal less illuminating: interviews with 6,000 dog owners who are recalling stuff.

Never mind that most of these same folks cannot recall what they had for dinner last night!

And never mind that the "problems" being recalled may say less about the breed involved than it does about the unthinking nature of people and their bumbling actions around dogs. A Chihuahua has a good reason to fear the kind of clumsy handling that a Great Dane might scarcely notice.

But never mind that; the problem must be down the leash and not up!

But, wait; there's more. It turns out this research is not based on a random sample of dog owners. Instead, this study is based on a survey of "11 American Kennel Club recognized national breed clubs" and "an online survey" of 101 questions, of which 26 questions were actually on canine aggression.

What? You mean only 11 breed clubs were contacted, and the rest was done with an "online survey?"

An online survey of who? The press articles do not say.

And it only gets better. Upon inspection, it turns out that the "data" here is so thin that broad statements are being made based on reports about -- literally -- a handful of dogs.

For example, Dachshunds are deemed to be psycho based on a survey that encompassed only 68 dogs.

When your "N" is this small, only a few offending dogs are needed to tip an entire breed into the "aggressive" zone.

And so it goes down the list. Only 56 Chihuahuas were in the survey. Only 78 Jack Russell terriers. Only 63 Beagles.

In fact, of the top four breeds indicted as "the most aggressive" this distinction was based on the supposedly bad temperament of about 30 dogs.

Not 30 dogs per breed, but 30 dogs across all four breeds!!

More troublesome still is that response bias does not seem to have been factored into the survey at all.

What is response bias? Simple: In this case, it would be the tendency of owners of large-breed dogs beset by "breed ban" legislation to downplay bites and aggression, and the tendency of the owners of small dogs to overstate how "tough" their little dogs actually are.

"Tricky-Woo is soooo fierce," says the hair dresser. "You should have seen him growling at that big ol' mastiff the other day."

Along with response bias, you may also have selection bias, which is to say that folks who pony up to take a 101-question survey about dog temperament may be more likely to have a disconnect between the dog they thought they were getting, and the dog they actually got.

Which is a nice way of saying that a small dog is just about as much work as a large dog, but the expectations of a small dog owner may be very different.

What freaks out the ill-informed owner of Chihuahua or the casual purchaser of a Jack Russell terrier may not raise the eyebrows of a Pit Bull or Mastiff-owner who comes expecting to have some problems, and who is more likely to to make training Job One from Day One.

Of course, none of this matters to newspaper reporters looking to score a quick headline.

Your average reporter got an "A" in English and failed math. An R-square value? What's that mean? A meaningful sample size? A survey response rate? Sample bias? Survey bias? And what do you mean when you say "correlation is not causality?"

Reporters know little about such things, and care even less if this kind of "detail" gets in the way of a quick headline.

"Attack of the Killer Dachshunds!"

That's the kind of headline that sells, and so we get unending amounts of junk science consumed like junk food by a gullible public more interested in entertainment than information.

Just imagine if, instead of dogs, this was a survey of adolescent aggression.

Can you imagine any self-respecting university doing a study of "human aggression by race" in which a handful of "ethnic-pride clubs" were questioned along with folks responding to an online survey about how their kids acted?

Woooo-eeee!
Now imagine that folks were simply asked to "recall" incidents of aggression or fear or excitability or attention-seeking by their children.

No objective critera or hard incidents of bad actions were needed; it was all about parental impression, not about arrest records or bullet holes. And, of course, the numbers of each "race" surveyed would never be large enough to tease out any meaningful data on such variables as age of mother, preschool attendance, or education.

And now imagine that when questions were asked and the answers coded, that murder, robbery, and rape were treated the same as a playground brawl, a shoving match, and yelling at the kids next door to get out of your back yard.


Would you consider that meaningful and important science?

Or would you call it something else?

The University of Pennsylvania should be well and truly ashamed of this shoddy bit of duff.