Reminder: Login to access new features and members-only content!

Register to be a member of our community. Its easy!

Register a new account

Already a member?

Log In here!


Did you find our content interesting or helpful? Help support the IPFD enhance health, well-being and welfare for dogs everywhere.

Jump to content
  • entries
  • comments
  • views

The sampling effect


Viewed: 5,409 times

I remember in the early months of my PhD to have assisted with a seminar by Raymond Coppinger on dog behaviour. At the beginning of his talk, he told the assembly that all comments and contributions would be welcome, providing the participant did not begin with “my dog did…”.  Over the last years, I have begun to think that this comment from Coppinger is the perfect illustration of one the biggest misunderstandings between dog breeders/owner and scientists, that I would call the sampling effect.


Coppinger's meaning was the following: you cannot make a generalization for the whole species, based on the sole example of your own dog. I think this applies not only to behaviour, but also to a lot of traits related to health.


With complex traits, to be able to make generalizations, you need to analyse data on a very large number of dogs… I will take the effect of inbreeding on longevity as an example. How many people have I heard saying “I have known many dogs with high inbreeding that have lived a healthy and long life”?

Well, guess what, they are surely right…




The preceding figure plots the longevity and inbreeding from more than 6,000 Britanny Spaniel (data from Leroy et al. 2015). This plot is not really impressive by itself.  If you consider the five dogs on the right side, with inbreeding close to 40% (remembering that 25% corresponds to a mating between full-sibs), one dog died at the age of 9 years, and four died between 15 and 17 years, the average longevity being around 11 years.


So, based on this example, people might conclude that after all, if they can see such inbred dogs live, for most of them, such a long life, then maybe inbreeding does not affect longevity.


Well, guess what, they would be completely wrong…


Wrong, because of the sampling effect, wrong because when considering complex phenomenon such as inbreeding and longevity, you cannot generalize from a small number of examples. You need a lot of data, not on a few dozen… You need at least several hundred, and if possible several thousand.  





Applying a simple linear regression of inbreeding on longevity to the same data, the impact is clear: for each 10% increase in inbreeding, longevity decreases by about 1 year. More elaborate models, taking into account the specific distribution of data, breeder or animal effect, confirmed this negative impact (Leroy et al. 2015). But the thing is that, as illustrated by the two figures, longevity is affected by a large number of other factors, related to genetics or environment. This is why it requires a lot of dogs to have an estimation of the global effect of inbreeding on the trait.  The individuals cited above were inbred and lived relatively long lives.  But they must be considered outliers, in this example, and their experience is not enough to impact the findings for the whole population.


By providing this example, my intent is not to criticize the expertise of dog breeders. The message would be more to underline the importance of confronting and sharing one's own experience, to be open to contradictory opinions, and try to gather data and analyse results in all objectivity. In the dog world, we still have a lot of work to do in that extent.



Leroy, G., Phocas, F., Hedan, B., Verrier, E., & Rognon, X. (2015). Inbreeding impact on litter size and survival in selected canine breeds. The Veterinary Journal, 203(1), 74-78.

1 Comment

Recommended Comments

Gregoire describes very typical and widespread human behaviour.  The evidence that smoking is bad for human health is irrefutable.  But bring that up in conversation and someone will say 'My uncle has smoked 3 packs a day since he was 12 years old and he's 90 years old.  Never hurt him!'.  The citing of an exception does not deny the rule.  When I have presented data from the Agria breed statistics and updates on the early mortality and poor survival of, e.g. Bernese Mountain Dogs, I am greeted by comments like 'My last bitch was healthy to 11 years'.  She certainly may have been; but it does not take away from statistics on thousands of dogs over many years.  


As Gregoire also suggests, the lack of strong evidence on representative populations leads to us relying on personal experience and anecdote rather than valid information for estimates of prevalence and risk.  And beyond the scarcity of good information, there is a lot of misinformation in the veterinary and lay literature, mostly, one hopes, out of misunderstanding rather than an intent to deceive.  Nonetheless, in the absence of clear results from large studies, discussions on contentious issues are generally unproductive when each side quotes anecdotal evidence to support their assertions, while accusing the other of misinterpreting the data.


I echo Gregoire's encouragement - let's keep an open mind, especially where definitive data are not available.  We must also realize that many of the tough issues of dog breeding, health, well-being and welfare are personal and emotional.  We must be extra-vigilant that we do not get blinding by our attachments to theories, history or personal preference.  However, the veterinary and scientific community should do more to provide the information that is needed to address these problems; and to provide it with appropriate and clear messages of the benefits and limitations of the data being provided, for there are no perfect studies that apply to all populations or individuals.  Our partners Agria Insurance and Vet Compass are doing what they can to add to our knowledge base on population statistics (see, e.g.: our section

on Data from Veterinarians | Insurance Data.  We are seeing more and more good work out of our KC partners as well, as they strive to produce quantitative studies from their large databases (e.g.: our section on  Estimated Breeding Values - Hip Dysplasia and Elbow Dysplasia and other papers cited in Hot Topics: Who's Talking About What?.


As always, thanks Gregoire for a thought-provoking blog.

Link to comment
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Blog Disclaimer
    The contents of this blog are for informational purposes only and represent the opinion of the author(s), and not that of the International Partnership for Dogs (IPFD). This is not intended to be a substitute for professional, expert or veterinarian advice, diagnosis, or treatment. We do not recommend or endorse any specific tests, providers, products, procedures, opinions, or other information that may be mentioned on, or linked to from this blog.

  • Create New...

Important Information

By using this site, you agree to our Terms of Use.