The need for theory

hmmm, that doesn't rhyme quite as well. Ben Bolker brought the following quote from Efron and Tibshirani (1986; "Bootstrap methods for standard errors ...") to my attention:

An important theme of what follows is the substitution of computing power for theoretical analysis. This is not an argument against theory, of course, only against unnecessary theory.

I've often thought of the need for theory as falling along a continuum of 1/n, so when your sample size is small you need strong theory to make predictions, and when large you can get away with less theory. In either case it helps if your theory is well tested in other cases, or you risk making predictions that are completely bogus.

Wolf Management reprise

On The Wildlife Society Blog Michael Hutchins criticized Deborah Peter's article in the Huffington Post on the current wolf harvest. One section in particular emphasizes why wolf management will be political, not scientific, and thus not a good candidate for AM:

I hate the fact that Congress intervened in the ESA with regard to wolf management. Management and conservation should be in the hands of scientists and professional managers and not in the hands of politicians. But why did this happen? Precisely because extreme animal rights proponents (and some extreme environmentalists)–unwilling to acknowledge that wolves have indeed recovered, pushed things too far, arguing for no control what-so-ever.

The reason it is political is precisely because different groups hold different values for wolves - ranchers vs. cool headed wildlife scientists vs. extreme animal rights proponents. Last time I looked, people are allowed to have different values, and when they do, politics, not science, will carry the day.

Making Decisions is hard!

Yes! making decisions is hard, and it saps brain energy, which in turn reduces self control! Eat chocolate before crossing the Rubicon!

Info gap uncertainty

You can't imagine how dreadfully unhappy I was to discover that not all uncertainty could be handled with probability, even subjective probability. My (former) student Max Post van der Burg wrote a paper on one approach to handling this type of uncertainty in structured population models, using the info-gap terminology developed by Yakov Ben-Haim. Yakov describes the approach in a book, which is both a bit expensive and a bit long for the casual reader. Lots of stuff in there! However, recently Yakov joined the blogosphere with tidbits intended to introduce his ideas in smaller doses.

Schooling one's thoughts

Quite a while back Jim Peterson (now at Oregon State), started me thinking about similarities and differences between approaches to Adaptive Management. One of my students, Jamie McFadden, took on this idea and conducted a small review of published AM studies, which is now available as "Evaluating the Efficacy of Adaptive Management Approaches: Is There a Formula For Success?". In it, Jamie outlines the attributes of AM projects that fall into two camps: the Experimental Resilience camp and the Decision-Theoretic camp. Jamie found that projects in the DT camp were steadily increasing in number, and that they tended to reach a higher level of success - as she defined it. I hope this article stimulates some broader conversation about what AM is and isn't, how to measure success, and how we can continue to improve - I believe it is time to "Adaptively Manage" Adaptive Management.

Conceptual models

Kate Buneau of Pacific Northwest National Laboratories sent the following link:

Cool Conceptual Model Diagram

Way complicated, but I love the way the different links light up when you mouse over a node. Positive and negative influences indicated with different links and symbols where the link reaches its target.

I'm reminded of a quote that Stephen Pacala gave in his talk - I can't remember the exact wording - something like the danger of building a complex model of a complex system risks having two things you don't understand - the model and the real system.

Horn tooting

One of the things I've been interested in for quite a while is making decisions with poor or no information - what social scientists since Keynes and Knight call uncertainty, meaning that there are no probability distributions available for the outcomes. If we're being honest with ourselves, this characterizes alot of circumstances when dealing with endangered species management. In such circumstances, one possible response is to "satisfice" rather than optimize the management actions.

Earlier this year Max Post van der Burg and I published an article in Ecological Applications Integrating Info-gap Decision Theory With Robust Population Management: A Case Study Using The Mountain Plover" where we used a combination of methods borrowed from robust control theory and satisficing to understand the value of a particular management action to a threatened species. This was a piece of Max's dissertation, and as usual in such things, he did all the hard and important work!

The core idea of "satisficing" is to find a decision that performs good enough, but over the largest possible number of ways of being wrong. In contrast, optimisation focuses on maximizing performance assuming that the system is perfectly understood - i.e. all the parameters are known perfectly and the system model is exactly correct - circumstances that are never true even in the best of times. So an optimal decision will usually outperform a satisficing decision if one's knowledge of the system is perfect, the satisficing decision will continue to do well even if the system model and its parameters are incorrect.

Of course, it is possible that a satisficing strategy is also the optimal strategy, and then we're happiest, but this doesn't seem to happen very often.

Max's contribution was to couple a matrix population model of Mountain Plover with the idea of satisficing to look at how well "nest marking" of Plovers performs as a conservation strategy. The upshot is that even if we are not sure about the life history of this species, nest marking increases the range of "wrongness" under which we will see positive population growth. What we didn't do was evaluate different types of actions against each other - this could easily be done, but was beyond the scope of what we wanted to achieve in the paper.