DS7002 Spatial Data Analysis
Practical Exercises
1
Session 9 – Near Things More Related Than Distance Things in London
(Spatial Dependence, Spatial Regression and Surface Analysis
with GeoDa and QGIS)
Objectives: To analyse spatial dependence / correlation for geographic profiling and
reasoning. To analyse spatial regression for analytic prediction modelling. To interpolate
point events and analyse raster data with surface techniques.
Task 1: Spatial Dependence / Correlation of Crimes and Deprivation in London
Problem solving: Measuring, mapping and analysing spatial dependence / correlation
of crimes and deprivation to support policing strategies
Functionality: Moran’s I in GeoDa
Data set: Crime variables and Multiple Index of Deprivation by London Borough
Input the Shape file of “london_life_polygon.shp” in GeoDa, and create a weights file
for calculating Moran’s I:
Geoda -> Tools -> Weights -> Create (suggest to select Rook Contiguity)
On “london_life_polygon.shp”, measure spatial dependence with following
functionalities by variables of ‘DEPRIV’ (score of deprivation), ‘CRIME_1000’ (total
counts of crime per thousand persons), ‘BURGLARY’ (counts of burglary per thousand
persons), ‘CRIM_DAMAG’ (counts of crime damage per thousand persons),
‘DRUG_OFFEN’ (counts of drug offence per thousand persons), ‘VEHICLES’ (counts of
vehicle theft per thousand persons), ‘VIOLENCE’ (counts of violence crime per
thousand persons).
GeoDa -> Space -> Univariate Moran’s I
-> Univariate Local Moran’s I
Does deprivation has strong spatial dependence? Which type of crime has
highest spatial dependence? What does the Univariate Moran’s I imply for
deprivation, total counts of crime and counts of burglary at the geographic
scale of borough? Copy Moran Scatter Plot and Cluster Map of ‘DEPRIV’,
‘CRIME_1000’ and ‘BURGLARY’ to your portfolio.
On “london_life_polygon.shp”, measure spatial correlation with following
functionalities by the pair of ‘DEPRIV’ & ‘CRIME_1000’ and the pair of ‘DEPRIV’ &
‘BURGLARY’.
GeoDa -> Space -> Bivariate Moran’s I
-> Bivariate Local Moran’s I
What does the Bivariate Moran’s I imply for the pair of ‘DEPRIV’ and
CRIME_1000’ and the pair of ‘DEPRIV’ and ‘BURGLARY’ at the geographic scale
of borough? Copy Moran Scatter Plots and Cluster Maps to your portfolio.
DS7002 Spatial Data Analysis
Practical Exercises
2
On “london_life_polygon.shp”, measure smoothed (or adjusted) spatial dependence
with following functionalities by rates of ‘BURGLARY’ (event variable) / ‘CRIME_1000’
(base variable), and ‘DRUG_OFFEN’ (event variable) / ‘CRIME_1000’ (base variable).
GeoDa -> Space -> Moran’s I with EB Rate
-> Local Moran’s I with EB Rate
Can you see any difference in results for ‘BURGLARY’ and ‘DRUG_OFFEN’ from
Moran’s I with EB Rate and Univariate Moran’s I ?
Try “london_ward_met_per1000.shp” with Univariate Local Moran’s I. by variables of
‘Total’ (total counts of crime per thousand persons) and ‘Burglary’ (counts of crime
burglary per thousand persons).
Find out any difference in spatial dependence between borough-level and
ward-level caused by MAUP. Copy Moran Scatter Plots and Cluster Maps to
your portfolio.
Try “london_pop20112016_plg.shp”, measure spatial-temporal dependence with
following functionalities by the population estimates in 2011and 2016.
GeoDa -> Space -> Differential Moran’s I
-> Differential Local Moran’s I
What does the Differential Moran’s I imply for temporal change of population
at the geographic scale of borough? Copy Moran Scatter Plots and Cluster
Maps to your portfolio.
Task 2: Spatial Regression of Health Indicators in London
Problem solving: Modelling spatial Regression of health indicators to support
healthcare strategies
Functionality: Regression in GeoDa
Data set: Helath variables and Multiple Index of Deprivation by London Borough
On “london_life_polygon.shp”, run regression modelling with following functionality
by ‘LIFE_MALE’ (life expectancy of male residents) as Dependent Variable and
‘DEPRIV’ (score of deprivation), ‘P_SMOKE’ (population percentage of smokers),
‘P_BINGE’ (population percentage of binge drinkers), ‘P_OBESE’ (population
percentage of obesity) and ‘P_HEALTHY’ (population percentage of healthy residents)
as Covariates. Please try both Classic and Spatial Lag models.
GeoDa -> Regression
Copy regression reports to your portfolio.
Try Session9_regression.r and copy results of regression modelling to your
portfolio. Please compare results from GeoDa and R.
DS7002 Spatial Data Analysis
Practical Exercises
3
Task 3: Population Surface in Haringey
Problem solving: Develop surfaces of social variables with interpolation to compare
or integrate data at different spatial scales
Functionality: interpolation in QGIS
Data set: Population centroids of LSOA in Haringey
Add “Population_Haringey_LSOA2011PWC.shp” and “haringey_boundary.shp”. Click
the button of “Zoom Full”.
QGIS -> Raster -> Interpolation
(In the window of Interpolation, Select Inverse Distance Weighting (IDW) as
Interpolation method. Choose “Distance Coefficient” as 4, click “Set to current
extent”, take default numbers of columns and rows as 300×300)
QGIS -> Raster -> Extraction -> Clipper
QGIS -> Layer -> Properties -> Style
(right click of mouse -> Properties -> Style -> Render type -> Singleband
pseudocolor, then change the legend)
(could be choose 5 classes)
QGIS -> Raster -> Extraction -> Contour
(try interval of contour lines as 100)
Try interpolation with different methods and parameters. Is interpolation
sensitive to methods and parameter settings? (just try, don’t need to copy to
portfolio)
Task 4: Terrain Surface around Waterloo Bridge
Problem solving: Explore natural surfaces with remote sensing techniques to provide
information for spatial data analysis
Functionality: Terrain Analysis in QGIS
Data set: LiDA DSM around Waterloo Bridge
Add “TQ3080_DSM_2M.asc”, Analysis the DSM with following functionalities:

-> Hillshade
-> Relief
DS7002 Spatial Data Analysis
Practical Exercises
4
Please export specified Scatter Plots and Cluster Maps from Task 1 (three variables for
Univariate Local Moran’s I, one pair of variables for Bivariate Local Moran’s I, one variable for
Local Moran’s I with EB Rate, one variable for “london_ward_met_per1000.shp” with
Univariate Local Moran’s I, the population estimates in 2011and 2016 for Differential Local
Moran’s I), and put them into your portfolio.
Please copy results of regression modelling from Task 2, export the interpolated raster
image from Task 3, and put them into your portfolio.
(QGIS: Project -> Save as Images, GeoDa: right click of you mouse -> Save Image as)