Idea Regional Regression
Our Approach: Use a separate regression function for different
regions.
Problem: Need to find regions with a strong relationship between
the dependent and independent variable.
Problems to be solved:
1. Discovering the Regions
2. Extracting Regional Regression
Functions
3. Develop a method to select which
regression function to use for a
new object to be predicted.
Source: http://www2.cs.uh.edu/~ceick/kdd/CE09.pdf
Motivation
Regional Knowledge & Coefficient Estimates
In geo-referenced dataset, most relationships only exist at
regional level but not at the global level.
1st law of geography: “Everything is related to everything else
but nearby things are more related than distant things” (Tobler)
Coefficient estimates in geo-referenced datasets spatially
vary we need regression methods to discover regional
coefficient estimates that captures underlying structure of
data.
Using human-made boundaries (zip code etc.) is not good
idea since they do not reflect patterns in spatially variance.
Motivation
Geo-Regression Analysis Methods
Regression Trees
Data is split in a top-down approach using a greedy
algorithm; uses constants as regression functions
Discovers only rectangular shapes
Geographically Weighted Regression (GWR)
an instance-based, local spatial statistical technique used
to analyze spatial non-stationarity.
generates a separate regression for each possible query
point “online”determined using a grid or kernel
a weight assigned to each observation that is based on its
distance to the query point.
Motivation
Example 1: Why We Need Regional Knowledge?
Arsenic
Fluoride
Regression Result: A positive linear regression line
(Arsenic increases with increasing Fluoride concentration)
Motivation
Example 1: Why We Need Regional Knowledge?
Location 1
Location 2
Arsenic
Fluoride
A negative linear Regression line in both locations
(Arsenic decreases with increasing Fluoride concentration)
A reflection of Simpson’s paradox[16].
Motivation
Example 2: Houston House Price Estimate
Dependent variable: House_Price
Independent variables: noOfRooms, squareFootage, yearBuilt,
havePool, attachedGarage, etc..
Motivation
Example 2: Houston House Price Estimate
Global Regression (OLS) produces the coefficient
estimates, R2 value, and error etc.. a model
This model assumes all areas have same coefficients
E.g. attribute havePool has a coefficient of +9,000
(~having a pool adds $9,000 to a house price)
In reality this changes. A house of $100K and a house of
$500K or different zip codes or locations.
Having a pool in a house in luxury areas is very different
(~$40K) than having a pool in a house in Suburbs(~$5K).
Motivation
Example 2: Houston House Price Estimate
Solution: To apply local regression to each zip code
produces 50+ sets of parameter estimates
it captures spatial variations in the relationship better than
global model
But it is very naïve and has problems
there is spatial variation within zip codes
assumes discontinuity but most spatial patterns are
continuous and they do not stop & start at the border.
Motivation
Example 2: Houston House Price Estimate
$350,000
$180,000
Houses A, B have very similar characteristics
OLS produces single parameter estimates for predictor variables
like noOfRooms, squareFootage, yearBuilt, etc
Motivation
Example 2: Houston House Price Estimate
If we use zip code as regions, they are in same region
If we use a grid structure
They are in different regions but
some houses similar to B (lake
view) are in same region with A and
this will effect coefficient estimate
More importantly, the house around
U-shape lake show similar pattern
and should be in the same region,
we miss important information.
