Research Proposal Outline

A research proposal for a simulation experiment is briefly describe the reasons for the planned experiment and offers a detailed experimental design along with a proposed data analysis. The outline below provides an initial structure for such a research proposal. While developing the proposal, keep in mind the subsequent reporting stage. Your research question and your anticipated reporting needs should guide your experimental design and proposed analysis.

Header

The document title should describe the research question. Include your full name, institutional information, and the date.

Research Question and Hypotheses

A research question inquires about the behavior of the simulation model. A hypothesis is a prediction about how changes in one or more model parameters affect a particular measured simulation outcome of the model. Therefore, your hypothesis must be linked to your output data. (See below.) An experimental design proposes a particular investigation of a research question, rooted in a simulation experiment.

  • State the general research question(s) addressed by this experiment.

  • State and justify one or more specific hypotheses, which your experiment will test. (A hypothesis usually proposes a causal relationship, stating how a specific output will respond to a specific kind parameter change.)

  • Briefly provide your current reasoning for each hypothesis that you offer. (Your reasoning may change after you run your experiment, but running the experiment does not happen during the planning stage.)

Model Choice

  • If you use or modify an existing model, specify this model so that any reader can easily locate it. Carefully cite the source of any borrowed code.

  • If your model is not identical to an existing model, give it a new name. You may include a version number or other more specific information.

  • If your model modifies an existing model, very briefly describe the changes. (You will eventually give a full description in a report.) If it is a brand new model, very briefly describe the model. (You will eventually give a full description in a report.)

  • Briefly explain why your simulation model is relevant to your research question.

Baseline and Treatments

  • Define the baseline scenario. This typically requires a table containing the name, type, baseline value, and source of each focal parameter. (The focal model parameters are the ones that the experiment varies.) The source may be a taken from the literature or an existing model, but when that is impossible it can also be an arbitrary reference point. Later, in the subsequent reporting stage, you may copy this table into your report.

  • You may wish to supplement the table of baseline parameters with a table listing any important exogenous constants. (These are important model parameters that you will not vary in any experiment.)

  • If you rely an existing model either directly or by modification, be sure to specify the relationship between your baseline and the original model. (The baseline scenario usually represents the baseline of the original model, or possibly a scenario considered in the original model.)

  • Provide a treatment table for this experiment. A treatment table gives a complete parameter-sweep description. Each parameter varied by the experiment must appear in a treatment table. A table row lists one parameter and describes the values used in the experiment. Make this table as concise as possible. (For example, if appropriate, you may state start, increment, and end values.) In your discussion of this table, state whether there is a particular reason for the choices of values. Create such a table even if you have only a single treatment variable, in which case the table will have only one row.

Randomness and Replicability

How many replicates will you run for each scenario? Give a reason for your choice. Describe how you will handle any stochasticity in the simulation model. How will your ensure that your results are fully replicable? For example, explain exactly how you will set the seed for the random number generator for each scenario and for each replicate of a scenario.

Input Data

Will there be any exogenous input data during a simulation? If so, describe them. (Many projects do not use any input data.) Describe the input data in enough detail that your results will be fully replicable. (This may involve providing the data in an appendix or in a stable internet repository.)

Output Data

Describe all of the values you intend to collect from your simulation experiment. (This is the simulation data for your analyses). Also, specify in detail how you collect these values. (For example, if experimental output from a NetLogo model will be collected via BehaviorSpace, say so, and be specific about the settings for the experiment.)

Data Analysis Tools and Methods

Describe your plans for the analysis of the data produced by this experiment. Propose a specific test of each hypothesis, based on your collected data. Rely on what you have learned about the design and analysis of experiments to plan an analysis that will help answer your research questions. Be very specific about your intended statistical analysis. For example, what kinds of statistical charts and tables do you plan to create, and how will they bear on your hypotheses? If you plan to run a regression, present the regression model that you currently plan to estimate, and relate it to your hypotheses.