Confirmatory Deliverable - Summer 2021 Final Project

Due: Tuesday, August 10th, 2021, 11:59PM Eastern Time [here]

Symposium: Thursday, August 12th, 2021 - 9AM - 12PM Eastern Time [gather.town] [space distribution]

Logistics

Step 1: Finalize Your Analyses

In this stage, you should not develop new analyses other than those that you have done in the exploratory deliverable. Pick one or two existing methods of analyses that you already have, address any concerns or take the next steps that your group plan to take to make your analysis more neat. Use your held out test set to evaluate your models/hypotheses. In general, try to make your analyses more rich, robust, and concise in this stage!

The feedback that we gave you for your exploratory deliverable should have given you a good idea of what changes you need to make to your project (if any). However, if you are still unclear about the next steps, feel free to come to Ellie’s Office Hours, or ask your TAs.

Common pitfalls/questions:

It’s easy for groups to focus too much on the model side of things - that is, questions such as: What is the best model to predict X? What is the best way to encode the object Y? While these questions are interesting, it strays from our focus on data: we want you to focus on things such as feature analysis (e.g, what features from your data would be the most meaningful to include in your models) and error analysis (e.g., what are the instances in which your model fail? what are the outliers that your models/analyses have to handle, and will your models/analyses be better if you remove the outliers)
For hypothesis testing analyses, even when your test set is small, you should still run your analyses on your test data. While the sample size may not be big enough to yield statistical significance, running your analyses on your test data will allow you to evaluate things such as whether a relationship remains positive/negative, or whether the effect size is similar to what you have if you run your analysis on the training dataset.

Step 2: Record Your Presentation

Prepare a recorded presentation (recommended 4 minute max, but no hard cap) that describes your final project - you can refer to the final project roadmap as a guideline. This video should not enumerate everything that you have done so far in the final project, but moreso tell a coherent story of your final project.

Particularly:

You probably should not try to cover all 3+ ways of analysis that you performed in the exploratory deliverable stage. Focus on one or two - those that you chose to complete the finalizing analysis step above. You can have a sentence or two about the analyses that you dropped (what you did, what the results were, and why you did not focus on them), but again: depth over distance.
This video will be presented to everyone else in the class who has little to no knowledge about what your project is about. Therefore, put on the hat of a salesperson giving an elevator pitch, and focus on telling a clear, accessible story that does not get bogged down by too many details. What the video presentation should focus on can be found in the roadmap.

As per usual, if you are ever unsure about anything, please reach out to Ellie or your TAs!

If you are unsure about how to record a presentation, we recommend checking this guide on how to record a presentation using Zoom, or this alternative tool.

After you have recorded your presentation, upload your presentation onto YouTube (making sure that the video is available to public), and hand in your presentation video link using the link at the top of this handout.

Data Science Final Project Symposium

Everyone is expected to participate in the final project symposium, where each group will have a dedicated space to themselves. Each group's presentation video will be embedded in the space, so anyone can come into the space, watch the video, and interact with members of the group to ask questions about the project.

All groups are expected to attend and participate for the entire duration, and each group should have at least one member in their space at all times to answer questions that people might have about their project. During the symposium, Ellie will walk around to watch each group's presentation video and ask each group questions (without any fixed schedule).

At the symposium, the following awards will be given out at the end:

Best Project Award: This award is given to the group that delivers the best final project overall.
Social Impact Award: This award is given to the group that is the most driven by social impact and/or has the most potential of delivering high social impact.
Presentation Award: This award is given to the group that gives the best presentation - think: visualization, or storytelling.
Innovative Award: This award is given to the group that includes the most innovative methods in their project.
Community Award: This award is given to individuals who contribute the most to the symposium - think: asking meaningful questions for other groups, helping explain core concepts to others in a beautiful way, or gassing other groups up.

Be sure to not miss out on these wonderful awards!

Evaluation

In addition to the criteria with which we evaluated your project in the previous deliverable, we will also evaluate your presentation by the extent to which your final presentation tells a good overarching story (as described in the final roadmap). The grade that we will give will be holistic and will be the final grade that you receive on the final project.