Managerial Control: Life with a High R2
Robert H. Giles, Jr. Department of Fisheries and Wildlife Sciences, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061-0321
Based on a lecture presented at the Wildlife Habitat Short course, April 11, 1989, Donaldson Brown Center for Continuing Education, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061-0321
People will mark from this very hour in 1989 the change in wildlife habitat management in North America.They will do so, or I will fail in my hour with you. I have no jokes, no warmup pattern. I am deadly serious. I see before me the leaders in wildlife management (of the USFS and related agencies), Then where shall we go together? What great things shall we build? I see you as a guerilla force infiltrating the ranks of your agency and doing some work there, some at the national level, some in regional shops. I see you influencing the way wildlife habitats are discussed, analyzed, and manipulated. I see you as controlling the outcome of the Earth forces. You are a control agent; you can be a more effective agent; you are gaining control.
To manage is to control. If you are a wildlife resource manager, you are controlling benefits produced and costs. You may not have all the control you desire, but you are in control. If you are not then I want a replacement! As a high-tax payer, I want returns on my money. I pay for control for desired change, for reduced risks, for reduced deviations.
You recall from basic algebra and statistics that this is a simple linear regression equation that depicts the line drawn. This is the fundamental model of habitat management. When I want more of Y (like bird food) then I add more of X (like fertilizer) and within the range I'll get more bird food.
These points are field observations signifying that when x had one value, y had a related value. The line drawn among one set of points (shown in B) may be very close to most of the points or distant as shown in C. You recall that the correlation R is poor in C and the R2 is very small. A perfect fit, all points being on the line, would result in an R2 of 1.0. In general, we all seek equations (and their lines) that match up well with the observed world.
We hardly ever get to work with only one factor, one x. It would be wonderful if we could get more or less of you symbolized as yby just changing x, thus changing y as suggested here.
y = a - bx - cz
The a is the height at the intercept and b and c are coefficients or the rates of change in x and z, the so-called independent variables. In the "realworld", x and z could be water and fertilizer that influences y which is bird food in a forest clearing.
The manager could change x and z to get a new y.
In SAS computer packages there is a concept called step-wise regression. It displays a multiple regression analysis but puts the variables out in the order of their contribution to this R2 measure of goodness of fit.
This picture (as others, once projected on transparent sheets) says, approximately, that if you really want to control y, change x10 and if you can afford it, change x2 and x1 but don't waste time on any of the others, no matter how many you dumped into the computer for analysis.
I view a multiple regression equation worked out for a management area as the control panel of an aircraft or space ship - a resource ship - and the manager can manipulate any or all. It is the simultaneous manipulation of many variables, the ecosystem knobs and switches, that is the job of the modern rational habitat manager.
R2 as you now recall, is an index to fit, an index to how well an equation accounts for variance, how well the points fall along the line, how well it "explains" y. For the manager it is how much control he or she is likely to have. If I gave you an opportunity with unlimited funds to change one variable
If I identified yas an index of successful, rational, high-net-benefit wildlife resource system and I listed this group, you before me, as an x variable, I would expect it to show up with a high R2.
Unless you miss the obvious and are prone to be a loner or have already said "to hell with this group," I suggest we all consider how we might live our professional lives with a high R2 .
Where are you in this equation? A dominant variable with control power or one far down the line in a step-wise analysis? Because you are here, I suspect you are inclined to increase your R2.
The needs are great, the field is awash with poor theory, competitive texts, insufficient research, enormous interests and demands, shifts from more game to less pest damage, and almost no leadership from the USFWS or the Wildlife Society. There is room for, high need for, a major controlling force - the force of this group; the force of you working within this very group.
Much too general? Perhaps. The pattern of thought just suggested can be useful. Use it, test it. Find the significant variables; work with them.
There is no time to play the game of "let's study everything" and then act.
Perhaps we can get more specific. I'll leave the very specific stuff to those in later sessions of this school. Let me suggest the rules of our guerilla force. These are the Green Rules, the G rules (for Green, Giles, and Guerilla).
Rule 1 - Advance the tentative deduction.
Given at least 200 dominant wildlife species in even the smallest Ranger District, it is impossible to ever study by classical and inductive techniques 250 factors about each species (as included in state data bases). If it took an average of one year to master each factor at a cost of $40,000 per factor (say half a scientist-year), then the total cost would be $2 billion by the time your great grandchildren could use the information on one Ranger District or wildlife management area.
There is not the time, talent, or funds to continue down the experimental pathway. It is irrational to do so; a pretense that some day we shall learn all of the basics if we keep at it long enough. Besides, there are not 200 species but 2000 for the serious ecologist!
Tentative is a world seldom heard in wildlife circles. It sounds insecure; wishy-washy.The need is to use it well. We know nothing; everything is uncertain. Even the theologian Martin Buber in "I-Thou" discussed "a holy insecurity."
The tentative deduction is a temporary suspension of disbelief, action, then feedback.
Rule 2 - Ignore the condition of insufficient time.
Basic research proceeds on the premise that a vast reservoir of knowledge is being filled and from it knowledge will some day be dipped. "Some day" is too uncertain; pressures are too great; delays are intolerable or too costly; decisions are to be made today and tomorrow, not someday.
The time line is no longer linear but exponential. Research must be highly applied and problem oriented. Let the "basic" be derived from the functional, applied and development work; waiting for the reverse to happen has not solved well wildlife resource management problems. We will never have enough time; efficiency experts will only improve our abilities by a few percent. Ignore time; decide. Then use feedback. The chances for making one big mistake are very small compared to the losses that accumulate in the years of indecision about hundreds of problem situations.
Rule 3 - Use the power of the computer.
Years ago (1960's) it was claimed that ecosystem problems were too big for the computer to handle.Now we have big computers with the capability and all thoughts seem focused on the microcomputer. There is great power available; it needs to be used. I presented plans for such use with Refuge System in 1969, and I still beat the same drum.
Planned modeling of small systems, even in undergraduate classes, can create large, functional, practical systems. Teams of experts could do so much more!
Rule 4 - Clarify objectives. Computer use, models built, should have clear objectives, answers to what do you really want to know from your model?
Until this type question is answered and the ones intimate to it - management for
then almost no successful habitat work can be done. At least we will not be able to decide whether successful or not.
It sounds silly, but if I had the objective of maximizing deer sightings on a District, I would drive bus loads of people past a deer in an enclosure. The objective rules the system.Once clearly and precisely stated, then creative efforts can be harnessed to achieve them Systems work to achieve objectives. Once made clear, specific and accounting done, including rewards and penalties, then everything can work and work increasingly well. Efforts spent on objectives are subversively useful.
Rule 5 - Manage life groups.
There is more difference between a wild turkey poult and an adult bird than between all of the species of thrushes or all of the species of spring warblers. Species are irrelevant, a topic for taxonomists. Life groups are the identifiable, managerial resource.
A beaver is one thing; a turkey needs management as two life groups; deer need management as fawn, buck, and antlerless creatures. These are not taxonomic units. We need to think of resource units. I call them "life groups."
"Featured species management," apparently stemming from my 1962 paper, is absurd. So is indicator species management. Both ignore the power of the computer. Both ignore life forms, both are static, both are decided by biologists, not the public.
Rule 6 - Let the clients rule.
On public lands, citizens are the clients. On farms, the owner is the client. The manager works for clients, achieving their objectives. The manager is subject to law such as "preserve species" but after that, benefits from resources are humanly assigned.
The benefits experienced by or desired by a manager may not and probably are not those of the client. (The probability of a match is close to zero.) When the manager becomes king or queen, then he or she is the client. Until then, the manager achieves the objectives of the client. As a member of the public he or she gets one vote, but that's all.
Rule 7 - The managers determine the methods.
Perhaps the clients rule but the managers provide the methods. The client states the objectives, demands, weights, etc., but the manager has the expertise to list alternative ways by which these may be achieved and to select among them.The manager separates objectives from practice. The difference is often obscured.
The public says: "do not use practice A." The public is saying (usually unknown to the public or the manager) "practice A in no apparent way achieves my objectives B1 B2 or B3 and reduces achievement of B4, B5, and B6. Don't use it." They may suggest an alternative (s), but for the practicing manager open to all inputs, these suggestions only enter a long list from which the optimum is selected - based on objectives.
Rule 8 - Manage the land cell.
We now can manage the land cell, some small tract, say about 1 to 3 acres, over large areas. These maps show the potentials - [GIS displays].
It is not science fiction, only laziness or misdirected funds not to imagine starting and forging ahead a system to manage cells.
Here are the points:
Rule 9 - Decide with models.
There are two major types of models: simulation (those that allow answers to "what if this change was made...?" and optimization (those that allow selection of best alternatives) . If Rule 4 (clarify objectives) is followed, then optimization will be used. I am really not interested except as a scientist in "what if...?" As a manager, I have a set of objectives and I can proceed to select the practice(s) that will best achieve them. The universities, full of scientists, have created in you the "what if ...?," "let's study the situation" thought pattern. You have to change this to the managerial "what is best ...?", or "what is optimum?" alternative pattern of thought.
Studies are needed but they should follow model building. From the model can be discovered what variables really need study; which ones most influence the performance measures of the system, and where will the greatest gains be made given the usual situation of low budgets, high risks, and insufficient time.
Rule 10 - Master the system, then the variables.
We now know enough to create realistic general models of most decision-related systems. I'm not talking about the biochemistry of isopods but about hunter trails, optimum rotations, and locating visitor centers. Decision-oriented systems we can model, but we rarely have done so. By modeling the system, by getting a rough, first-cut approximate system operation, the manager will then have the organization, the structure, the tests, the feedback, to advance progressively that model based on team work with other members of this group, the literature, and local research. A priori model building provides the direction for what studies to do, it provides immediate use of results of research, it allows coordination, and it provides healthful feedback to the researcher working along who needs, at least occasionally, that feeling of having achieved something - almost anything!
A priori model building can be called "planning" but I know few planners who would agree.
Rule 11 - Variance in samples can be explained.
No, variance is not mystical, not due to unknown cosmic forces. It can be reduced by studying the distributions of nature. Few are linear. At least segment your data. Study linear features within set ranges. At least, as a rule, try the transformation.
Z = log (x + 1 )
Also, use the curvilinear PC software now available. All variance cannot be explained; the physicists shifted from determination to a probabilistic world view years ago. (We continue to act as if it were deterministic.)
The major emphasis of Rule 11 is that there are dominant abiotic forces influencing our results. Many of these are readily available and need to be brought into multivariate studies to provide the statistical controls needed. The millions of dollars invested by public agencies can provide
Now that many of these are readily available, and more becoming more available, it seems reasonable to avoid saying "results could not be concluded to be significant because of the large variances encountered" but to control the variance.
Some people claim that abiotic factors are not of interest (personally by biologists) but because they cannot be manipulated. "You can't do anything about the weather." The rational manager will shift perspective and, for example, put plants where the factors are right for them. He or she did not change soil moisture on a site but in the management area selected sites where the soil moisture was appropriate for the plants. I perceive the manager as having great control over abiotic factors.
Rule 12 - Be confident in lour confidence.
How many significant figures should be used in the answer to
10 x 2.3684 x 1.49 + 169?
The numerical answer is 204.2892. The proper answer is 200. In an ecosystem of 100 variables, there may be many with amounts measured to several decimal places. When brought together with objectives (if they ever are , (and need to be), the objectives rarely are assigned weights of relative importance more than high, low, medium or on occasion units of relative importance of from 1 to 10.
These latter numbers require the end computation for the entire managerial system to be rounded to 1 significant figure! This awareness, the knowledge of what to do with a standard error of an estimate and general concepts of statistical variance, natural variability, and the expectation of catastrophe such as fire, flood, and epidemic, and the other natural catastrophe, the budget cut, -- all lead to the rational wildlife habitat manager working with confidence levels of about 75%. The gambling tables of the west and throughout the country succeed because they win more than they loose, that's all. They do loose! They're not greedy; they win when the odds are greater than 50%. After this, the calculations are based on volume.
All you have to do is win a little, occasionally, in the forest wildlife habitat game because so many acres are involved over so many years. Tests at the 95% level are absurd in our world. We want to be confident at high levels but do not forget the actions of the sophisticated engineers. They design bridges with well-measured and many-decimal equations -- then double the results for safety. They understand precision and probability and the associated costs of each. We need to advance with appropriate confidence.
If you follow these 12 G Rules as I try to do, then we can have a profound influence on wildlife habitat management. If you do not, then you have an obligation to explain why not, to correct and revise them, or advance for me your own, perhaps as a summary of this course.I am willing to listen, but do not interpret that as a sign of any insecurity about these rules. The wildlife resource is important to me personally.These are extremely difficult times, but there are new concepts and technology. What seems missing is the old fire, the desire to get moving, the concern for wildlife, the magic of yore.
I think that fire can be rekindled. The Rules give us some structure. Let's work together. You personally or we as a group may contribute abundantly to the R2 of the equation of how well the wildlife resource is now managed.
Perhaps you will share ideas with me about some of the topic(s) above .
Robert H. Giles, Jr.
July 3, 2005