not a member?   forgot your password?

Illuminating Statistical black boxes: promoting understanding and correct interpretation in support of forest management and policy analysis

This issue has two related, but distinct parts: statistical knowledge of natural resource professionals and the more specific question of understanding Type I and II error.

a) Statistical knowledge. Climate change, invasive species, fragmentation, plummeting economies, and many other threats (and opportunities) force foresters to make decisions about an uncertain future. Occasionally such decisions are based on faulty analyses due, in part, to a lack statistical rigor. Without correct interpretation of the results, foresters run into the risk of making wrong decisions that can have long-lasting consequences and can be costly. However, interest and participation in statistical training, sampling, analysis, and modeling has been steadily decreasing over the decade. Knowledge in these subject areas is decreasing at a time when our needs for solid understanding of them have never been greater. Possible vehicles for addressing this part of the issue include keynote speakers at the 2010 convention (for example, dynamic, engaging speakers like Tony Starfield and conversant statisticians who are good communicators-they are out there-why not have them take forestry cases and make the concepts relevant to our audience?), a special issue of the Journal of Forestry, and development of online tutorials in statistical application tailored to natural resource problems and hosted by SAF.

b) Understanding type I and II error. On those occasions when foresters tackle an issue scientifically, chances are there will be talk of "scientific proof", generally referencing the notion of statistical significance. However, it is not clear that there is widespread understanding of the values aspects of choosing alpha in a risk tradeoff with beta regarding the respective danger of a Type I (calling a difference real when it really is not) & Type II (calling a difference not real when in fact it is) error. This is alarming given that most of what we do involves just such risky decisions with costly and imperfect information, and there's a risk to our credibility when we talk in public about "science" without this understanding, as we may well be the only ones at the table with this understanding deficit. The profession could be very well served by re-exposure to these ideas from introductory statistics, organized around examples such as managing wildlife diseases, formulating climate policy, and funding research. Possible vehicles include an issue of the Journal of Forestry, or even a whole track on it at the 2010 convention, perhaps followed up by a monograph or other publication carefully targeted and packaged so as to promote understanding among our diverse membership.

Implied strands of this theme
Many/most foresters have forgotten or gone fuzzy concerning key statistical precepts upon which the scientific aspects of forestry are founded. Many of us have come to rely on a combination of our own personal experiences and observations built on not necessarily representative anecdotal evidence and an array of "priesthood" models and corporate datasets built by researchers, scientists and analysts but often with important details, limitations, and assumptions poorly understood by the managers and policy makers that rely upon them, often for purposes different than those for which such models and datasets were originally intended. As a result, decisions that are well supported by facts and scientific evidence are rarer than they should be, and the risks of decisions where uncertainties are inevitable are not well understood.

Implied questions

• What will it take for foresters to understand the significance and importance of this problem?
• How can foresters be engaged to invest the effort to tune up their understanding of statistics? How can the benefits be made plainly apparent, and how can we make this tuning up fun?
• What avenues can provide the best learning experiences most efficiently (and with the least pain on the part of the learners) for the greatest number of forestry professionals? How can the Society support the development of online statistical courses that will cover basic to intermediate to advanced concepts in a way that is relevant to foresters?
• How can we build a system for ongoing knowledge improvement and refreshment such that gains in statistical understanding are not one-time events followed again by decline?
• How can we ensure that sound statistical principles are incorporated in all operational forest inventories?
• How can we interpret the statistical significance of results obtained from remotely sensed data such as aerial photographs, satellite images and LIDAR and overlays of such geographic layers?
• Understanding the black box-what questions should all foresters ask about their models?

Planned actions so far

• A series of thought-provoking essays titled “A Series on Statistics and Science” are available on the SAF website.

Send your ideas for other actions to Jeremy Fried at saf at jeremyfried.net, FSTB Emerging Issues Committee Chair.