Individual Projects

Finding a Topic

Here's one approach:

1. Start with an existing model in a domain that interests you. The model should have free parameters that are estimated by fitting to data. The model should also be a process model, not just descriptive “curve fitting”.
2. Identify the fundamental phenomena that the model addresses.
3. Of those phenomena, which is the simplest to replicate empirically? Invent a reduced version that can be run in 10 minutes.
4. Simplify the model to its essential mechanisms and parameters necessary to address the phenomenon.

Alternatively, start with a robust phenomenon and then find a simple model of it.

Presentations: Number and Duration

• Two presentations per person: One regarding the empirical phenomenon; one regarding the model.
• Eighteen (18) minutes per presentation (i.e., four presentations per day).
• There will also be a final report and poster due at the Final Exam time (which is **).

Tentative Presentation Dates

• See Oncourse Announcement for who presents on each date.
• Experiments must be available to run by one week after your first presentation.
• Final report and poster at Final exam time (**).

What to Present

• Talk 1: Empirical Phenomenon
  • Motivation & background (answer the question, Who cares?)
  • Design (conditions, variables/factors)
  • Expected effect (differences between conditions)
  • Analysis (what kinds of statistical tests)
• Talk 2: Model
  • Intuitive cognitive principle or mechanism
  • Formal expression of that mechanism
  • Parameters and their interpretation
  • Qualitative predictions of the model (in same format as expected effect from experiment)
  • Measure of (mis-)fit to be used, and why
• Poster
  • Size: Approximately 3 ft tall by 4 ft wide. It does not need to be a single sheet of paper or cardboard; i.e., you can print it on several pieces of standard paper and arrange them within a 3x4 area.
  • Support: Either to be taped on wall with masking tape (not scotch tape) or on self-supporting cardboard display. (You provide the tape or the cardboard.)
  • Format: Use a very large font! It must be readable from a distance. Arrange the information in columns, not rows.
  • Keep it concise: The gist must be discernable in a mere 6 minutes! (This is good practice for posters at real conferences.) In the 120-minute period there will be three shifts of six posters, each shift being about 35 minutes long (plus 5 minute set-up time).
  • Content: Should include the following sections
    • Project title and your name and departmental affiliation
    • Abstract
    • Motivation
    • Design of experiment
    • Results and statistical analysis
    • Model description
    • Fit measure and procedure
    • Model best fit parameter values and predictions
    • Conclusion
    • References
• Final Report: Submit to Prof. Kruschke the following:
  • A copy of the poster.
  • Upload to Oncourse (a) your data analysis program(s) and (b) your model program well commented. Too many comments is far preferable to too few.

How to Prepare Talks

• Decide what essential points you need to make about each of the items in "What to Present" (above).
• Rehearse your presentation so that you realize what pre-requisite issues need to be explained for each of the essential points. It's difficult if you start making your presentation, try to make an essential point, and suddenly realize that you need to explain some background details before the essential point can be clear.

How to Present Talks

• Visual aids, such as PowerPoint or overhead transparencies, are recommended but not required. Chalk board and hand-outs are also an option, put these tend to be less easy for your audience to navigate.
  • Keep the overheads sparse; avoid too much text.
  • Use a large font (typically 24pt minimum).
• Speak loudly. (Imagine you're speaking only to the person at the far back of the room, and remember that the overhead projector and/or computer projector make a lot of noise.)
• You have only 18 minutes, so make your points concisely.