Facilitating Problem Framing in Project Plan Based Learning

Assignment Task

The project plan is a short document that describes the analysis you plan to carry out. It needs to contain sufficient detail so I can (i) evaluate its internal logic (i.e. the match between the real-life policy issue, its simplified framing as a solvable decision problem, and the data and functions you plan to use to solve it), and (ii) estimate its feasibility within the short timeframe available.

Reminder: the project is about trying to frame a decision problem such that it can be solved . This means that the project is an exercise in effective simplification, and it is not about getting the perfect answer. In the process, you will learn how challenging it can be to match ambitions and understanding of real-life complexity to models and available data in many analytical contexts, and how vulnerable recommendations can be to underlying assumptions and choices.

In line with the prospectus, the project plan needs to include short descriptions of the following items:

• What is the real-life policy issue you would like to help solve? What are its most essential characteristics?
• What are your decision variables ? Define the space of options from which the policy maker will choose. This can be a continuous variable (e.g. budget level), a set of binary variables (e.g. selected sites), a choice between different land cover maps (e.g. if you use Invest), or others. Start with a single dimension; you can always add more later.
• What is the objective function the decision maker is trying to maximize or minimize with their choices? Define it. iIdicate how you will measure it. If you have multiple objectives, decide whether you will model some of them as constraints (e.g. budget) or whether you plan to combine them into a single function (e.g. benefits minus cost).
• How will you model the simplified functional relationship between decision variables and your objective function? This could be anywhere from extremely simple (Lab 2: "if I protect this planning unit, all animals in it will survive") to a little more complex (Lab 3: "if I create an incentive program, some landowners will enroll, but not all; the creation of an easement will reduce the risk of forest loss to zero"). It will be easiest to start with a very simple relationship; you can always consider plausible alternatives.
• What datasets do you need and have to implement your analysis? If you don't build on top of an existing lab dataset, create a table of the datasets you would ideally want to have and the data you have access to. Indicate whether you have downloaded and inspected the data and assessed its suitability for your purpose. On the next page is an example of how such a table could look for Lab 2.

Project ideas 

There are a few different routes you can take to pick your final project. One is to identify a question based on a problem you are interested in – perhaps something that you have already thought about, but to which you want to add a quantitative spatial analysis.

Alternatively, your project can be a thoughtful extension of an exercise we do in the labs. To illustrate what that option means, here is a non-exhaustive list of examples. You're welcome to pick any of these, or to come up with different variations on the themes we cover in the lab.

Informing policy choices:

• Site selection (Marxan): which land parcels within the U.S. region of [state, region] should be protected if the goal was to conserve at least 10% of the habitat of each species listed in the State Wildlife Action Plan at the lowest cost? 
• Site selection (Marxan): How does the quality of conservation cost estimates affect species conservation priorities in the Colombian Andes? 
• Return-of-investment analysis: where should a donor with a fixed budget and interests in avoided carbon emissions invest within the U.S. region of [state, region] to maximize the additional amount of carbon that we expect to stay on the ground as the result of that choice?
• Return-of-investment analysis: what would be the set of locations that the Biden administration (or a selected city government) could turn into new urban parks in order to reduce racial inequalities in recreational access [in city X]

Exploring alternative methodological choices:

• Site selection (Marxan)
  • How different are the results of conservation planning studies if we use an exact optimizer (prioritzr in R) instead of simulated annealing (Marxan)? 
• Improving predictions
  • Will a different estimator (e.g., neural network) give us better predictions of land values in Massachusetts?
  • Will a different set of predictors (feature engineering) give us better predictions of land values in Massachusetts?Cross-validation
  • Which modeling strategy most accurately predicts land values in a larger region it hasn’t seen yet (i.e., spatial cross-validation instead of random cross-validation)? 
• Impact evaluation
  • How do the estimates of the impact of protected areas on forest cover change if we use novel and recently published matching algorithms (e.g., MALTS for Python) instead of Mahalanobis distance matching? 

This Management has been solved by our PhD Experts at My Uni Paper.