Week 12: Tools for Final Projects
PPOL 6805 / DSAN 6750: GIS for Spatial Data Science
Fall 2024
Class Sessions
Midterm Post-Mortem
- On the one hand: You will be able to “complete” missed parts and receive credit for “completing” them!
- On the other hand: You can also lean into spatial regression and use bonus HW to make up missed midterm points instead!
Choose Your Own Adventure
- Bonus HW5A: Re-doing Midterm Q3-Q5 (with new dataset)
- You’ve incurred “fixed cost” of understanding Q1-Q3…
- \(\implies\) Build on this fixed cost and complete ~Q4 + ~Q5 without time limit!
- Bonus HW5B: Spatial Regression
- Your final project uses areal rather than point data…
- \(\implies\) To prepare, do HW5B to learn/practice spatial regression and earn bonus points for doing so!
Final Project Details
- Presentations: 6:30-9pm, Wed, 4 December, 2024
- Reports: Due 5:59pm, Fri, 13 December, 2024
Presentation Setup
- Each of your desks becomes a “table” at a GIS conference!
- Everyone can go around and ask others about their projects 😻
- Once I have grilled you on yours (~2-5mins), you are free to leave
- But, FOOD \(\implies\) you can also stay!
Report Setup
- GitHub repository (for your portfolio!)
- GH Pages site:
your_username.github.io/gis_project
- GH Pages site:
- …What do you put in that GitHub repository?
- Quarto Manuscript
- What do you put in the Quarto manuscript?
- Writeup + Visualizations + Code, interspersed!
- “Literate Programming” \(\implies\) Reproducible Results!
Immediately-Relevant Tools for Final Projects!
- Research Methodology (Hypotheses)
- Visualizing Spatial Data
- Weighted Connection Matrices \(\mathbf{W}\)
- Remote-Sensed / Raster Data
- Data Anonymization / Synthetic Datasets
Research Methodology: Hypotheses
Social Science (McCourt):
Machine Learning (DSAN):
(Secretly My Opportunity to do More Spatial Regression!)
- What explains previously-observed instances of separatist insurgencies?
- Can we predict separatist insurgencies?
- One (spatial) idea: how far away are [centers of power] from [regions of countervailing power]?
- \(X\) = Distance from capital, \(Y\) = Insurgency
- Unit of observation: Regions? Insurgent uprisings? Countries?
Operationalizing
- The variables in previous slide are conceptual
- Operationalizing = “Turning into measurable quantities”
- \(X\) = MeanDistance(Capital, Insurgent Region), \(Y = \mathbf{1}[\text{Insurgency}]\)
- Alternative:
\[ Y = \begin{cases} 2 &\text{if Successful Insurgency} \\ 1 &\text{if Failed Insurgency} \\ 0 &\text{if No Insurgency} \end{cases} \]
Connection Matrices
spdep- Neighbors if Centroids are close
- Neighbors if Capitals are close
- 🤔
Visualizing Spatial Data
- My recommendation for final presentations: Leaflet
Remote-Sensed / Raster Data
- You’ve seen (to some extent)
terra stars: Same group behindsf- Google solar panel data
.tiffiles: Dynamically loaded
Anonymization / Synthetic Datasets
- Differential privacy
- Used by the US Census(!) Since 2020