Week 1 Notes - Course Introduction

Key Concepts Learned

spatial autocorrelation of the errors

1. random errors (good errors):

• No systematic pattern
• Scattered across space
• Prediction equally good everywhere
• Model captures key relationships

2. Cluster errors (bad)

• Spatial pattern visible
• Under/over-predict in areas
• Model misses something about location
• Need more spatial features!

spatial lag of errors - spatial lag = averrage value of neighbors - create the spatial lag: my error vs avg error of 5r nearest neighbors

Moran’s I - +1 = Perfect positive correlation (clustering) - 0 = Random spatial pattern - -1 = Perfect negative correlation (dispersion)

• neighbors:

1. Contiguity: Polygons that share a border; Queen vs. Rook
2. Distance: All within X meters; Fixed threshold
3. k-Nearest” Closest k points; Adaptive distance

spatial lag/error models - no spatial lag models for prediction: simultaneity problem; prediction paradox; data leakage in cv

Coding Techniques

library: spdep

Questions & Challenges

• Which file I should go to when I make changes
• The whole process of making changes

Connections to Policy

• Upload my work to my portfolio for visualization

Reflection

• How different platform can connect and work with each other
• I want to practice more and dig deeper