Monday, July 23, 2012

Regenerative Ecology: A Look Ahead

Over the upcoming months I hope to address a few related topics on this blog. Since I expect to be fairly busy, and I feel the internet is already a dumping ground with loads of low-quality information, my goal is going to be quality over quantity. But I wanted to give a sense of what this project is supposed to be all about, assuming yours truly keeps himself on track :)

1) The origins, ecology, and evolution of grasslands and ruminant animals. I will briefly cover the relevant biology of grasses and ruminant animals, and then discuss the changing extent and nature of grasslands over the ages, and their present significance. Since grasslands account for 25-33% of the earth's terrestrial surface, it's important to understand the basics of how they function, contribute to biodiversity, and provide invaluable "ecosystem services" to our species.

2) The role of ruminant livestock in agriculture from a national and global perspective. I will consider how ruminant animals contribute to nutrient cycling on organic farms, the maintenance and health of grasslands and rangelands, and the human food supply. We will consider the common environmentalist objections to ruminant animals. In particular, I will focus on the methane production of ruminant animals: Why does it happen? How much of it is released? How does livestock husbandry affect methane production?  What are the global sources and sinks?

3) A detailed look at the global carbon cycle, including a layman's primer on ecosystem carbon storage, and other sequestration strategies. The goal is to understand how land management can contribute to climate change mitigation and adaptation.

4) Exploration of the subject of resilience and sustainability. What does it mean for landscapes, and ecosystems? For human economies? How does politics help or hinder? What can we do to foster resilience in our own lives and communities? I will argue against many of the common assumptions about sustainability, green development, permaculture and so on.

                                                                      ----------

So what is "regenerative ecology" anyways?

I suggested a partial answer in my first post. It wasn't a very good explanation though. I think that the answer to this question will become clear as I address the above issues. Nevertheless, a few more words may be in order.

Our civilization has been under-written by relatively cheap, abundant fossil energy since the industrial revolution. I take this to be axiomatic. At the same time, the legacy of colonialism, excesses of capitalism, foolish decision-making and other politico-economic forces have encouraged exploitation rather than wise stewardship of natural resources and capital. So far, we've done a good job at insulating ever growing numbers of people from the condition of ecosystems on which we all depend, but this too has been facilitated by cheap energy.

 Consider agriculture.Yields have risen, even as natural capital has literally been eroding into the sea:
-Global agriculture produces several tons of eroded topsoil for every ton of food produced
-Industrialized intensified agriculture (such as in the US), uses between 5-12 kcals of fossil energy for every kcal of food energy yielded
-Global agriculture has become a net sink of energy, rather than a net producer 

The consequences of this pattern of development are numerous:
-Despite record yields, more and more people go hungry
-Nutrient export into oceans creates hypoxic zones
-Loss of carbon from topsoils depletes long-term fertility and soil quality, and contributes to global warming
-Human health suffers from the over-production commodity crops for the processed food industry

At the same time, the outlook for global energy supplies appears tight, to put it mildly. The consequence is rising prices for everything.

We can do things differently! We can rely on ecosystem services, and sunlight energy. Again, agriculture is the most tractable example. It should
-build soil fertility rather than deplete it
-conserve or restore biodiversity rather than imperil it
- be a net yielder of energy, rather than a net sink

I don't want to be glib about this transition. Indeed, that's why I'm writing this blog. But if we want to rely increasingly on natural ecosystem processes and agroecosystems for our necessities and goods, we cannot just talk about sustaining them; in many cases, we have to regenerate them! This is analogous to politics, where we need to eschew "lesser of two evil" choices, and start supporting things that are actually part of the "solution matrix".

This means that excellent land stewardship has to become the norm, and not the exception.

So, broadly speaking, I define "regenerative ecology" as the application of ecological sciences towards the aim of regenerating ecosystem health while returning a yield to their human inhabitants and managers.







Friday, July 20, 2012

Bayesian Modeling the LA Veterans PUFA Study

(UPDATE: edits added to the model section, correcting bonehead mistake. There is no stratification of model families by treatment, because the treatment is a categorical variable in the model... ) 

Introduction

There's been an interesting (to dorks like me) discussion going on over at Chris Masterjohn's blog "Mother Nature Obeyed" concerning statistical frameworks, and the meaning of that ever-elusive p-value:

 http://www.westonaprice.org/blogs/cmasterjohn/2012/07/17/im-95-confident-this-is-a-good-definition-of-a-p-value/

Specifically, we've been discussing the peculiar results of one of the first human diet trials which substituted modern processed seed oils, for traditional animal fats-- the LA Veterans Administration trial. This trial is often cited in reviews and meta-analyses to adduce support for the hypothesis that polyunsaturated-fat-rich vegetable oils reduce the risk of cardiovascular mortality. In fact, as Masterjohn and others have pointed out, the actual results are more interesting.

(More details can be found here: http://www.westonaprice.org/know-your-fats/good-fats-bad-fats-separating-fact-from-fiction)


The Study

There were two treatment groups- an intervention that substituted vegetable oils, and a control that stayed with traditional animal fats (e.g. butter). Significantly, the study population had a mean age above 60 years old, and was carried out for a relatively long time, 8 years.  The researchers tracked total mortality, cardiovascular mortality,  and cancer mortality. The cancer data were only possible because of the age of the cohorts, and this is evidently a unique aspect of this study.

In brief, the total mortality data were a wash, the cardiovascular mortality showed a benefit to vegetable oils, but the cancer data showed a benefit to animal fats!

But the plot thickens. It turns out the randomization was not as effective as it should have been at controlling for smoking rates, so the animal fat group was saddled up with a much larger rate of heavy and moderate smokers. In essence, this has lead Masterjohn to (quite reasonably) conclude that, absent confounding, the animal fats cohort would probably have not had a higher CHD risk, and would have had an even lower relative cancer risk than was observed!

The study authors concluded that the effects of treatment (animal fat versus vegetable oil) on cancer risk "only" attained a p-value of 0.06 (above the arbitrary threshold of p=0.05 for avoiding type I error, i.e. incorrectly rejecting a true null hypothesis), and was thus "not significant". But the effect size was decent (20% or so), and there's that pesky issue of confounding, confound it!

So, the question is, how seriously should we take this p-value? Are there alternative approaches to exploring these data that avoid the pesky arbitrariness?

This lead to a discussion about interval estimation, it's relation to p-values, and the differing interpretations that frequentists and Bayesians bring to the table.

 P-values and Confidence Intervals

In brief, frequentist statistical hypothesis testing is all about rejecting or failing to reject a null hypothesis. It asks for the probability of the data given the null hypothesis P(D/Ho). The null hypothesis specifies no statistical correlation (slope=0 in a linear regression, for example). Rejecting the null hypothesis doesn't mean that your alternative hypotheses are necessarily true. Frequentist intervals ("confidence intervals") are constructed so that a p% interval, in a long-run of hypothetical experiments, contains the "true" parameter value p% of the time.  It is incorrect to say that you are p% certain that particular interval contains the true parameter value, since it either does or doesn't. This is hugely counter-intuitive.

In contrast, Bayesian hypothesis testing first specifies a prior (either informative or non-informative), then calculates a posterior distribution or probability for a specific hypothesis, P(Hi/D). This is a more direct assessment of some particular quantitative hypothesis than the frequentist framework of rejecting a null hypothesis. It is correct to assert a p% confidence in the hypothesis, given the posterior data. Likewise, Bayesian intervals (credibility index, credible set, etc.) are expressions of confidence that the underlying parameter (which Bayesians treat as a random variable, and not fixed) has a value in that range *based on these data* and not a hypothetical long-run of experiments. 

It is very important to note that the choice of a non-informative prior usually makes a Bayesian analysis basically identical to it's frequentist counter-part. It can do all the same things, but is often computationally clunky, and there are still philosophical differences in interpretation, as suggested above.

In the end, I don't think these differences in approach mean much for how to interpret the p-value of 0.06 the study authors reported. Since this study is unique, there's little justification for a Bayesian informative prior to re-run an ANOVA style hypothesis test with.

But I think there's another way (although I'm admittedly way out of my depth in speculative land). My background is in ecology and agriculture, not medicine or nutrition, although the fields interest me.

Bayesian Modeling- An Alternative Proposal

 In the comments section of the above post (http://www.westonaprice.org/blogs/cmasterjohn/2012/07/17/im-95-confident-this-is-a-good-definition-of-a-p-value/) I outlined an alternative approach to these data. Rather than quibbling with the p-value the study authors reported, I would build a Bayesian statistical model and use Bayesian model selection to select among the best, and would estimate credible sets for the parameter values. This approach is not an alternative to hypothesis testing (which was already done), but is a complementary approach to understanding and getting information out of data.

I wrote:
"I would use a Bayesian statistical model that has both informative and uninformative priors (be warned that I’m way out of my depth here, my only familiarity with Bayesian inference is in ecological studies):

I would build a generalized linear model with three predictor variables: veggie-oil/animal-fat(categorical), age (categorical, probably) and smoking rates (continuous). The response variable is cancer rate. You have two sets of models (one for each treatment).

Rate=B(0)+B(1)(Veggie-oil OR animal fat)+B(2)(smoking)+B(3)(age)+interaction terms

You could use frequentist model selection and parameter estimation at this point. However, using a Bayesian approach you would specify priors for each parameter. Since it seems there should be robust data for the effects of smoking and age on cancer risk, it makes sense to incorporate them as Bayesian priors (rather than treating this study as a de novo universe for cancer risk attributable to smoking and age). Might the NIH have such data tables?

I would choose a non-informative Bayesian prior for the animal-fat/veggie-oil parameter. This is where the absence of previous studies comes into the picture.

Anyways, I would evaluate these data with Bayesian model selection and generate credible sets for the parameters. Rather than simple hypothesis testing we’ll: 1) select the “best” models for cancer risk, and 2) build credible sets for the parameters of interest to explore their biological meaning.

P-values would never enter into this analysis. Afterall, we already have a p-value (p=0.06), and it’s arbitrary to say that’s not significant but p=0.05 is! I think this statistical model would be far more interesting."

Model Selection and Interpretation

Using information-theoretic methods (Bayesian analogues of AIC), we'll discover what the best models are for predicting cancer risk in these cohorts. If none of the best models include a particular parameter, it's probably safe to say that parameter isn't really meaningful in this study.  Also, it would be interesting to note if any of the best models contain interaction terms, like, say (oil)X(age).

The credible sets around the parameter will tell us what the range in effect size is, and since we're being Bayesian we're p% sure of it!

Finally, the advantage of the Bayesian approach is that we should be able to integrate high-quality data from, say, the NIH on the influence of smoking rates and age on cancer for cohorts such as were in the LA Veterans study, and not pretend that this study is an isolated universe for sampling cancer risk factors.

For the vegetable oil or animal fat parameters, we'd use an uninformative prior, because there's no justification for anything else. Nevertheless, the results of our model (the posterior for that factors) could become priors for future studies of similar design...

This is the beauty and the danger of Bayesian priors.

Conclusions

Before doing the modelling, I'm not sure what to expect. It may be some of my assumptions are unwarranted. The devil is always in the details.  At any rate, this kind of modeling is an interesting complement to ANOVA-style hypothesis testing.

My probably over-hasty thought is that these data would, in model selection, show a protective effect for the animal fats. The inference would certainly be controversial! I'm certain that no-one in Harvard Public Health would believe it ;)




















Tuesday, July 17, 2012

Is Red Meat Risky? Complexity and Context in Human Health and Diets


Note: I wrote this one up in immediate reaction last March to Harvard's "red meat is risky" study. I bring up a number of issues in this post that have interested me for some time.

The recent "Harvard Meat Study" purports to show that red meat consumption is risky. By this, the researchers mean that a broad questionaire-based assessment of two study populations-- the Health Professionals Follow-Up Study and the Nurses Health Study-- showed a higher all-cause mortality for those including red meat in their diets (interestingly, the risk was even higher for those consuming processed red meats). Here’s the meat of the study (pun intended):

A combined 23,926 deaths were documented in the two studies, of which 5,910 were from CVD and 9,464 from cancer. Regular consumption of red meat, particularly processed red meat, was associated with increased mortality risk. One daily serving of unprocessed red meat (about the size of a deck of cards) was associated with a 13% increased risk of mortality, and one daily serving of processed red meat (one hot dog or two slices of bacon) was associated with a 20% increased risk.
Among specific causes, the corresponding increases in risk were 18% and 21% for cardiovascular mortality, and 10% and 16% for cancer mortality. These analyses took into account chronic disease risk factors such as age, body mass index, physical activity, family history of heart disease, or major cancers.”

Correlation is not causation. This is a mantra that is well worth repeating. Although it may be overdone in the blogosphere, it really is worth thinking through carefully and repeatedly. What studies like this do is establish a statistical correlation (including an effect size and some level of statistical significance). This is not a demonstration of biological causation. In fact, there is no clear causation here at all. This is a trade-off in large scale epidemiology- mechanisms and pathways are almost necessarily obscure1.

The task is to decide what these statistics mean. In this case, we are talking about a large class of food. We have therefore to deal with the complexity of the food itself, its role in human nutrition, and the whole complex of behaviors that go into people eating. The researchers hope they have accounted for a sufficient array of these factors in order to isolate the effect of the variable of interest, in this case red meat consumption per se. 

I have not seen the data, and am not familiar with the particular methods that are used in epidemiological studies of this sort. So I cannot comment on how well they performed this aspect of analysis. Being a large-ish team from Harvard, I’m sure they got the statistics technically correct. For instance, they had to “correct” for the fact that the red meat group was more obese and smoked more than the rest. If there is a non-linearity at work in the system dynamics, then the very premise of correcting for such “competing” risk factors to isolate the risk associated with red meat per se, is flawed2.

In epidemiology there are other strategies as well. The gold standard in medicine is a randomized double-blind intervention trial that follows a treatment and a control group for some period of time and assesses outcomes. This is the only way to be sure that we are avoiding the pitfalls of the illusory world of “data-mining”. If the treatment groups are well chosen and the intervention specific enough, we can be relatively certain in attributing a causal connection between intervention and outcome.

In contrast, this study is an observational study, albeit a prospective cohort which is definitely stronger than, say a retrospective cohort. The mortality effect described in this study would be a lot more convincing if this were an intervention trial that eliminated red meat in one group, allowed a certain consumption in another, and held the rest of the diets constant, or at least had a treatment and control group with a broadly representative (and normally distributed) range of baseline diets.

The problem is that in a complex system, there may be pathways of causality that are non-linear, and/or may be distributed statistically in unexpected ways. For instance, there may be a sub-set of people who are “hyper-responders” to red meat, and experience a dramatic increase in mortality, while everyone else is fine, or even has a lower mortality3. The net result, averaged over the whole population, may show up as a slight increase in risk, such as this study described. Also, a confluence of two or more synergistic, negative, factors may be necessary for the ill effect of red meat to manifest-- for instance obesity and smoking. I’m sure the researchers must have looked for such “interaction effects”, but when the number of variables and possible pathways is high, it may not be possible with ordinary methods.

Another problem is that this study did not bother to distinguish factory-farmed, conventional-grain-finished red meat, from grass-fed meat. We already know that the two kinds of foods differ significantly as to composition and the balance of health-promoting factors. The “null hypothesis” has got to be that they have different effects in human nutrition. In the end, this study may be simply a referendum on conventional feedlot beef.

Similarly, cooking and preparation were not assessed. For instance, people often eat red meat off the grill, replete with char and smoke carcinogens. Now, I love me some tasty grilled beef, but I’m not going to suggest that it’s healthy!

The effects of a given food are complex in themselves, and also mediated by dietary context4. So it may be that an even more stringent interpretation is necessary: in the context of a SAD (standard American diet--which is hypercaloric, hyper-processed, high in artificials and low in protective nutrients), the consumption of conventional feedlot beef and its processed derivatives, is associated with a slightly increased risk in mortality. However, the pathways and mechanisms of this potential toxicity are unknown, and until a suitable intervention trial is completed, we won’t really know if this effect is real. Extrapolations to naturally grass-finished beef and a non-SAD dietary context are, at this stage, flights in fancy.








1 My point is not to trash talk epidemiology. For one thing, it’s a hugely useful field of study and we need it. For another, I’m not expert in it. My main trope is the need, in thinking about human health and nutrition, to weigh these kinds of studies in the context of other high-quality sources of information; things that may be true in the statistical aggregate, may not be useful guidelines for any particular person or definite group.
2 Depending on the nature of the non-linearity, actually. The question is whether the researchers can correctly identify the form of non-linearity and parametrize it in their statistical model. In a system of high complexity, this is unlikely.
3 Dr. Ronald Krauss is actually working on the angle of iron levels. It may be that excess heme-iron, and iron storage overload, underlie most of the negative effects. Like many other things, red meat may be good in moderation, and problematic in excess- and what defines moderation may be very individual.
4 Context is super important in assessing diets because the addition of one food generally has to come at the exclusion of another. So, in choosing to eat one thing in particular, you are choosing *not to eat* a whole bunch of other alternatives. Also, the nutritive quality of foods in part depends on what other foods are eaten, and partly also the organism’s needs at any given time. Finally, if we consider foods as a complex mixture of nutritive, anti-nutritive and toxic factors, the overall dietary composition is irreducibly important. The usefulness of a food in human nutrition has to be assessed in this context, and cannot be predicted from its own constituents alone. An example, tannins from coffee, tea, and other plant foods may reduce iron absorption- mitigating risk of overload in those who are susceptible.






Monday, July 16, 2012

Introduction

Welcome to my blog! I've avoided starting one for quite some time now, but I really wanted to have a space to collect my thoughts on issues I'm passionate about. The name "regenerative ecology" reflects both a philosophical tenet, and a call to action. In short, I believe that environmentalism as a whole is entering a "Third Wave", embracing the role of human agency and management in shaping landscapes and ecosystems. We can no longer pretend that human decision-making is not a key driver of our collective environmental health or decline, and has been for quite some time. Moreover, in many human-managed landscapes- forests, rangelands, agroecosystems- we can indeed maintain or improve ecological integrity whilst simultaneously providing for human needs. We must avoid false dichotomies at all costs.

At the same time, our global environmental crises are reaching a tipping point. Global climate change is only one of the more visible, and worrisome manifestations of the widespread malfunction we have no choice but to try and mend. Unfortunately, there is also a fair degree of complacency that is emerging in the discourses of "sustainability", "green development", "local foods" and so on. At least in the United States, I see a troubling amount of ennui amongst the so-called progressives, a willingness to be satisfied with half-measures and "lesser of two evils" choices. So "regenerative ecology" refers also to the work to be done. 

Finally, I want this blog to stay grounded in the personal, concrete and goal-oriented present. Although some posts will be philosophical, and others technical in nature, a theme I want to return to frequently is how to incorporate what we've learned into our daily lives, and to discern what can and should be done at the manageable scale of household and community. Right now, our political system is deeply broken- and our movement needs to grow in spite of the politicians and the systemic corruption, not wait around for "enlightened government".

Anyhow, I will be elaborating and clarifying all this much more in subsequent posts.