Spatial Analysis and Modeling

(GIST 4302/5302)
Guofeng Cao
Department of Geosciences
Texas Tech University
 Outline of This Week
• Last week, we learned:
– spatial point pattern analysis (PPA)
– focus on location distribution of ‘events’
• This week, we will learn:
– spatial autocorrelation
– global measures of spatial autocorrelation
– local measure of spatial autocorrelation
 Spatial Autocorrelation
• Tobler’s first law of geography
• Spatial auto/cross correlation

If like values If there is no apparent If like values tend

tend to cluster relationship between to be located
together, attribute value and away from each
location then there is other, then there
then the field
zero spatial
exhibits is negative
autocorrelation
high positive spatial
spatial autocorrelation
3
 Spatial Autocorrelation
• Spatial autocorrelationship is everywhere
– Spatial point pattern
• K, F, G functions
• Kernel functions
– Areal/lattice (this topic)
– Geostatistical data (next topic)

4
Spatial Autocorrelation of Areal
Data

5
Positive spatial autocorrelation
2002 population
- high values density

surrounded by nearby high values

- intermediate values surrounded
by nearby intermediate values
- low values surrounded by
nearby low values

6
Source: Ron Briggs of UT Dallas
Negative spatial autocorrelation competition for space

- high values
surrounded by nearby low values Grocery store density

- intermediate values surrounded

by nearby intermediate values
- low values surrounded by
nearby high values

Source: Ron Briggs of UT Dallas 7

 Spatial Weight Matrix
• Core concept in statistical analysis of areal data
• Two steps involved:
– define which relationships between observations are to
be given a nonzero weight, i.e., define spatial
neighbors
– assign weights to the neighbors
• Making the neighbors and weights is not easy as
it seems to be
– Which states are near Texas?

8
 Spatial Neighbors
• Contiguity-based neighbors
– Zone i and j are neighbors if zone i is contiguity or
adjacent to zone j
– But what constitutes contiguity?
• Distance-based neighbors
– Zone i and j are neighbors if the distance between them
are less than the threshold distance
– But what distance do we use?

9
Contiguity-based Spatial Neighbors
• Sharing a border or boundary
– Rook: sharing a border
– Queen: sharing a border or a point

rook queen Hexagons Irregular

Which use?
10
 Example

Source: Bivand and Pebesma and Gomez-Rubio

11
 Higher-Order Contiguity

1st
order

Nearest
neighbor

rook hexagon queen

2nd
order
Next
nearest
neighbor

12
 Distance-based Neighbors
• How to measure distance between
polygons?
• Distance metrics
– 2D Cartesian distance (projected data)
– 3D spherical distance/great-circle distance
(lat/long data)
• Haversine formula

13
 Distance-based Neighbors
• k-nearest neighbors

Source: Bivand and Pebesma and Gomez-Rubio

14
 Distance-based Neighbors
• thresh-hold distance (buffer)

Source: Bivand and Pebesma and Gomez-Rubio

15
Neighbor/Connectivity
Histogram

Source: Bivand and Pebesma and Gomez-Rubio

16
 Side Note: Box-plot
• Help indicate the degree of dispersion and
skewness and identify outliers
– Non-parametric
– 25%, 50%, 75% percentiles
– end of the hinge could mean
differently depending on implementation
– Points outside the range are usually
taken as outliers

17
 Spatial Weight Matrix
• Spatial weights can be seen as a list of
weights indexed by a list of neighbors
• If zone j is not a neighbor of zone i, weights
Wij will set to zero
– The weight matrix can be
illustrated as an image
– Sparse matrix

18
A Simple Example for Rook case
• Matrix contains a:
– 1 if share a border
– 0 if do not share a border

4 areal units 4x4 matrix

A B C D
A B A 0 1 1 0
W =B 1 0 0 1
C D C 1 0 0 1
Common border D 0 1 1 0
19
20
 Sparse Contiguity Matrix for US States -- obtained from Anselin's web site (see powerpoint for link)
Name Fips Ncount N1 N2 N3 N4 N5 N6 N7 N8
Alabama 1 4 28 13 12 47
Arizona 4 5 35 8 49 6 32
Arkansas 5 6 22 28 48 47 40 29
California 6 3 4 32 41
Colorado 8 7 35 4 20 40 31 49 56
Connecticut 9 3 44 36 25
Delaware 10 3 24 42 34
District of Columbia 11 2 51 24
Florida 12 2 13 1
Georgia 13 5 12 45 37 1 47
Idaho 16 6 32 41 56 49 30 53
Illinois 17 5 29 21 18 55 19
Indiana 18 4 26 21 17 39
Iowa 19 6 29 31 17 55 27 46
Kansas 20 4 40 29 31 8
Kentucky 21 7 47 29 18 39 54 51 17
Louisiana 22 3 28 48 5
Maine 23 1 33
Maryland 24 5 51 10 54 42 11
Massachusetts 25 5 44 9 36 50 33
Michigan 26 3 18 39 55
Minnesota 27 4 19 55 46 38
Mississippi 28 4 22 5 1 47
Missouri 29 8 5 40 17 21 47 20 19 31
Montana 30 4 16 56 38 46
Nebraska 31 6 29 20 8 19 56 46
Nevada 32 5 6 4 49 16 41
New Hampshire 33 3 25 23 50
New Jersey 34 3 10 36 42
New Mexico 35 5 48 40 8 4 49
New York 36 5 34 9 42 50 25
North Carolina 37 4 45 13 47 51
North Dakota 38 3 46 27 30
Ohio 39 5 26 21 54 42 18
Oklahoma 40 6 5 35 48 29 20 8
Oregon 41 4 6 32 16 53
Pennsylvania 42 6 24 54 10 39 36 34
Rhode Island 44 2 25 9
South Carolina 45 2 13 37
South Dakota 46 6 56 27 19 31 38 30
Tennessee 47 8 5 28 1 37 13 51 21 29
Texas 48 4 22 5 35 40
Utah 49 6 4 8 35 56 32 16
Vermont 50 3 36 25 33
Virginia 51 6 47 37 24 54 11 21
Washington 53 2 41 16
West Virginia 54 5 51 21 24 39 42
Wisconsin 55 4 26 17 19 27
Wyoming 56 6 49 16 31 8 46 30 21
 Style of Spatial Weight Matrix
• Row
– a weight of unity for each neighbor relationship
• Row standardization
– Symmetry not guaranteed
– can be interpreted as allowing the calculation of
average values across neighbors
• General spatial weights based on distances

22
 Row vs. Row standardization

A B C Divide each
number by the
row sum
D E F
Total number of neighbors Row standardized
--some have more than others --usually use this
Row
Row
A B C D E F Sum
A B C D E F Sum
A 0.0 0.5 0.0 0.5 0.0 0.0 1
A 0 1 0 1 0 0 2
B 0.3 0.0 0.3 0.0 0.3 0.0 1
B 1 0 1 0 1 0 3
C 0.0 0.5 0.0 0.0 0.0 0.5 1
C 0 1 0 0 0 1 2
D 1 0 0 0 1 0 2 D 0.5 0.0 0.0 0.0 0.5 0.0 1
E 0 1 0 1 0 1 3 E 0.0 0.3 0.0 0.3 0.0 0.3 1
F 0 0 1 0 1 0 2 F 0.0 0.0 0.5 0.0 0.5 0.0 1
23
 General Spatial Weights Based on
Distance
• Decay functions of distance
– Most common choice is the inverse (reciprocal) of the distance
between locations i and j (wij = 1/dij)
– Other functions also used
• inverse of squared distance (wij =1/dij2), or
• negative exponential (wij = e-d or wij = e-d2)

24
25
 Example
• Compare three different weight matrix in
images

26
Measure of Spatial
Autocorrelation

27
 Global Measures and Local Measures
• Global Measures
– A single value which applies to the entire data set
• The same pattern or process occurs over the entire
geographic area
• An average for the entire area
• Local Measures
– A value calculated for each observation unit
• Different patterns or processes may occur in different
parts of the region
• A unique number for each location
• Global measures usually can be decomposed
into a combination of local measures
28
 Global Measures and Local Measures
• Global Measures
– Join Count
– Moran’s I, Geary’s C, Getis-Ord’s G
• Local Measures
– Local Moran’s I , Geary’s C, Getis-Ord’s G

29
Join (or Joint or Joins) Count Statistic

– 60 for Rook Case

– 110 for Queen Case
30
 Join Count: Test Statistic
Test Statistic given by: Z= Observed - Expected
SD of Expected
Expected = random pattern generated by tossing a coin in each cell.
Expected given by: Standard Deviation of Expected (standard error) given by:

Where: k is the total number of joins (neighbors)

pB is the expected proportion Black, if random
pW is the expected proportion White
m is calculated from k according to:

31
 Gore/Bush Presidential Election 2000

Actual
Jbb 60
Jgg 21
Jbg 28
Total 109 32
 Join Count Statistic for Gore/Bush 2000 by State

Actual Expected Stan Dev Z-score

candidates probability Jbb 60 27.125 8.667 3.7930
Bush 0.49885 Jgg 21 27.375 8.704 -0.7325
Jbg 28 54.500 5.220 -5.0763
Gore 0.50115 Total 109 109.000

• The expected number of joins is calculated based on the proportion of votes each
received in the election (for Bush = 109*.499*.499=27.125)
• There are far more Bush/Bush joins (actual = 60) than would be expected (27)
• Positive autocorrelation
• There are far fewer Bush/Gore joins (actual = 28) than would be expected (54)
• Positive autocorrelation
• No strong clustering evidence for Gore (actual = 21 slightly less than 27.375)

33
 Moran’s I
• The most common measure of Spatial Autocorrelation
• Use for points or polygons
– Join Count statistic only for polygons
• Use for a continuous variable (any value)
– Join Count statistic only for binary variable (1,0)

Patrick Alfred Pierce Moran (1917-1988)

34
 Formula for Moran’s I
n n
N  w ij (x i  x)(x j  x)
i 1 j1
I n n n
( w ij ) (x i  x) 2
i 1 j1 i 1

• Where:
N is the number of observations (points or polygons)
x is the mean of the variable
Xi is the variable value at a particular location
Xj is the variable value at another location
Wij is a weight indexing location of i relative to j 35
 Moran’s I
• Expectation of Moran’s I under no spatial
autocorrelation
E(I)  1/ (N1)

• Variance of Moran’s is complex and exact

equation is given at textbook d&G&L
• [-1, 1]

36
 Moran’s I and Correlation Coefficient

• Correlation Coefficient [-1, 1]

– Relationship between two different variables
• Moran’s I [-1, 1]
– Spatial autocorrelation and often involves one (spatially indexed)
variable only
– Correlation between observations of a spatial variable at location
X and “spatial lag” of X formed by averaging all the observation
at neighbors of X
 n Correlation
1(yi 1
i  y)(x i  x)/n Coefficient
Note the similarity of the
n n

 (y  (x
numerator (top) to the measures
i  y) 2
i  x) 2
of spatial association discussed
earlier if we view Yi as being the
i 1 i 1 Xi for the neighboring polygon
n n
(see next slide)
n n n n
n n
N  w ij (x i  x)(x j  x)
i 1 j1
w
i 1 j1
ij (x i  x)(x j  x)/  w ij
i 1 j1
n n n
( w ij ) (x i  x) 2
= n n

 i   i 
i 1 j1 i 1
2 2
(x x) (x x)
Spatial i 1 i 1
auto-correlation n n 38

Source: Ron Briggs of UT Dallas

 n Correlation
1(y
i 1
i  y)(x i  x)/n
Coefficient
n n

 (yi  y)
i 1
2
 (x i  x)
i 1
2 Spatial
weights

n n
Yi is the Xi for the
n n n n

w (x i  x)(x j  x)/  w ij

neighboring polygon
n n ij
N  w ij (x i  x)(x j  x)
i 1 j1
i 1 j1 i 1 j1
n n n =
( w ij ) (x i  x) 2
n n
i 1 j1 i 1

 i
(x
i 1
 x) 2
 i
(x 
i 1
x) 2

Moran’s I n n 39

Source: Ron Briggs of UT Dallas

Statistical Significance Tests for Moran’s I
• Based on the normal frequency distribution with

Where: I is the calculated value for Moran’s I

I  E(I ) from the sample
Z
Serror( I ) E(I) is the expected value if random
S is the standard error
• Statistical significance test
– Monte Carlo test, as we did for spatial pattern analysis
– Permutation test
• Non-parametric
• Data-driven, no assumption of the data
• Implemented in GeoDa

40
 Moran Scatter Plots
We can draw a scatter diagram between these two variables (in
standardized form): X and lag-X (or W_X)

The slope of this regression line is

Moran’s I
41
Moran Scatter Plots

High/High
Low/High positive SA
negative SA

Low/Low High/Low
positive SA negative SA

42
Moran Scatterplot: Example

43
 Moran’s I for rate-based data
• Moran’s I is often calculated for rates, such as crime
rates (e.g. number of crimes per 1,000 population) or
infant mortality rates (e.g. number of deaths per 1,000
births)
• An adjustment should be made, especially if the
denominator in the rate (population or number of births)
varies greatly (as it usually does)
• Adjustment is know as the EB adjustment:
– see Assuncao-Reis Empirical Bayes Standardization
Statistics in Medicine, 1999
• GeoDA software includes an option for this adjustment

44
 Geary’s C
• Calculation is similar to Moran’s I,
– For Moran, the cross-product is based on the deviations from the mean
for the two location values
– For Geary, the cross-product uses the actual values themselves at each
location
– Covariance vs. variogram

n n
N  w ij (x i  x j ) 2
i 1 j1
C n n n
2( w ij ) (x i  x) 2
i 1 j1 i 1

45
 Geary’s C vs. Moran’s I
• Interpretation is very different, essentially the opposite!
Geary’s C varies on a scale from 0 to 2
– 0 indicates perfect positive autocorrelation/clustered
– 1 indicates no autocorrelation/random
– 2 indicates perfect negative autocorrelation/dispersed
• Can convert to a -/+1 scale by: calculating C* = 1 – C
• Morain’s I is more often used

46
Statistical Significance Tests for Geary’s C
• Similar to Moran
• Again, based on the normal frequency distribution with
C  E (C ) Where: C is the calculated value for Geary’s C
Z
Serror( I ) from the sample
E(C) is the expected value if no
autocorrelation
S is the standard error

however, E(C) = 1

47
 Hot Spots and Cold Spots
• What is a hot spot?
– A place where high values
cluster together
• What is a cold spot?
– A place where low values
cluster together

• Moran’s I and Geary’s C cannot distinguish them

• They only indicate clustering
• Cannot tell if these are hot spots, cold spots, or both

48
 Getis-Ord General/Global G-Statistic
• The G statistic distinguishes between hot spots and cold spots. It
identifies spatial concentrations.
– G is relatively large if high values cluster together
– G is relatively low if low values cluster together
• The General G statistic is interpreted relative to its expected value
– The value for which there is no spatial association
– G > (larger than) expected value  potential “hot spots”
– G < (smaller than) expected value  potential “cold spots”
• A Z test statistic is used to test if the difference is statistically
significant
• Calculation of G based on a neighborhood distance within which
cluster is expected to occur

Getis, A. and Ord, J.K. (1992) The analysis of spatial association by use of
distance statistics Geographical Analysis, 24(3) 189-206
49
Formulae of General G

50
 Comments on General G
• General G will not show negative spatial autocorrelation
• Should only be calculated for ratio scale data
– data with a “natural” zero such as crime rates, birth rates
• Although it was defined using a contiguity (0,1) weights
matrix, any type of spatial weights matrix can be used
– ArcGIS gives multiple options
• There are two global versions: G and G*
– G does not include the value of Xi itself, only “neighborhood”
values
– G* includes Xi as well as “neighborhood” values
• Significance test on General G and G* follows the
similar procedure as used for Moran’s I
51
 Local Measures of
Spatial Autocorrelation

52
Local Indicators of Spatial Association (LISA)
• Local versions of Moran’s I, Geary’s C, and the Getis-
Ord G statistic
• Moran’s I is most commonly used, and the local version
is often called Anselin’s LISA, or just LISA

See:
Luc Anselin 1995 Local Indicators of Spatial
Association-LISA Geographical Analysis 27: 93-115

53
Local Indicators of Spatial Association (LISA)

• The statistic is calculated for each areal unit in the data

• For each polygon, the index is calculated based on neighboring
polygons with which it shares a border
• A measure is available for each polygon, these can be mapped
to indicate how spatial autocorrelation varies over the study
region
• Each index has an associated test statistic, we can also map
which of the polygons has a statistically significant relationship
with its neighbors, and show type of relationship

54
Example:

55
 Calculating Anselin’s LISA
• The local Moran statistic for areal unit i is:
I i  zi  wij z j
j

where zi is the original variable xi in z  xi  x

i
SDx
“standardized form”
or it can be in “deviation form” x  x
i
and wij is the spatial weight
The summation j is across each row i of the
spatial weights matrix.
An example follows
56
Contiguity Matrix 1 2 3 4 5 6 7
Code Anhui Zhejiang Jiangxi Jiangsu Henan Hubei Shanghai Sum Neighbors Illiteracy

Anhui 1 0 1 1 1 1 1 0 5 65432 14.49

Zhejiang 2 1 0 1 1 0 0 1 4 7431 9.36
Jiangxi 3 1 1 0 0 0 1 0 3 621 6.49
Jiangsu 4 1 1 0 0 0 0 1 3 721 8.05
Henan 5 1 0 0 0 0 1 0 2 61 7.36
Hubei 6 1 0 1 0 1 0 0 3 135 7.69
Shanghai 7 0 1 0 1 0 0 0 2 24 3.97

5
4
1
6 7
2
3
57

Source: Ron Briggs of UT Dallas

 Contiguity Matrix and
Row Standardized Spatial Weights Matrix
Contiguity Matrix 1 2 3 4 5 6 7
Code Anhui Zhejiang Jiangxi Jiangsu Henan Hubei Shanghai Sum

Anhui 1 0 1 1 1 1 1 0 5
Zhejiang 2 1 0 1 1 0 0 1 4
Jiangxi 3 1 1 0 0 0 1 0 3
Jiangsu 4 1 1 0 0 0 0 1 3
Henan 5 1 0 0 0 0 1 0 2
Hubei 6 1 0 1 0 1 0 0 3
Shanghai 7 0 1 0 1 0 0 0 2 1/3
Row Standardized Spatial Weights Matrix
Code Anhui Zhejiang Jiangxi Jiangsu Henan Hubei Shanghai Sum

Anhui 1 0.00 0.20 0.20 0.20 0.20 0.20 0.00 1

Zhejiang 2 0.25 0.00 0.25 0.25 0.00 0.00 0.25 1
Jiangxi 3 0.33 0.33 0.00 0.00 0.00 0.33 0.00 1
Jiangsu 4 0.33 0.33 0.00 0.00 0.00 0.00 0.33 1
Henan 5 0.50 0.00 0.00 0.00 0.00 0.50 0.00 1
Hubei 6 0.33 0.00 0.33 0.00 0.33 0.00 0.00 1
Shanghai 7 0.00 0.50 0.00 0.50 0.00 0.00 0.00 1

Source: Ron Briggs of UT Dallas 58

 Calculating standardized (z) scores
xi  x
Deviations from Mean and z scores.
X X-Xmean X-Mean2 z zi 
SDx
Anhui 14.49 6.29 39.55 2.101
Zhejiang 9.36 1.16 1.34 0.387
Jiangxi 6.49 (1.71) 2.93 (0.572)
Jiangsu 8.05 (0.15) 0.02 (0.051)
Henan 7.36 (0.84) 0.71 (0.281)
Hubei 7.69 (0.51) 0.26 (0.171)
Shanghai 3.97 (4.23) 17.90 (1.414)

Mean and Standard Deviation

Sum 57.41 0.00 62.71
Mean 57.41 / 7 = 8.20
Variance 62.71 / 7 = 8.96
SD √ 8.96 = 2.99
59

Source: Ron Briggs of UT Dallas

 Row Standardized Spatial Weights
Matrix
Calculating LISA
Code Anhui Zhejiang Jiangxi Jiangsu Henan Hubei Shanghai

Anhui 1 0.00 0.20 0.20 0.20 0.20 0.20 0.00

Zhejiang 2 0.25 0.00 0.25 0.25 0.00 0.00 0.25
Jiangxi 3 0.33 0.33 0.00 0.00 0.00 0.33 0.00
Jiangsu
Henan
4
5
0.33
0.50
0.33
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.50
0.33
0.00
wij
Hubei 6 0.33 0.00 0.33 0.00 0.33 0.00 0.00

I i  zi  wij z j
Shanghai 7 0.00 0.50 0.00 0.50 0.00 0.00 0.00
Z-Scores for row Province and its potential neighbors
Anhui Zhejiang Jiangxi Jiangsu Henan Hubei Shanghai
Zi
Anhui 2.101 2.101 0.387 (0.572) (0.051) (0.281) (0.171) (1.414) j
Zhejiang 0.387 2.101 0.387 (0.572) (0.051) (0.281) (0.171) (1.414)
Jiangxi
Jiangsu
(0.572)
(0.051)
2.101
2.101
0.387
0.387
(0.572)
(0.572)
(0.051)
(0.051)
(0.281)
(0.281)
(0.171)
(0.171)
(1.414)
(1.414)
zj
Henan (0.281) 2.101 0.387 (0.572) (0.051) (0.281) (0.171) (1.414)
Hubei (0.171) 2.101 0.387 (0.572) (0.051) (0.281) (0.171) (1.414)
Shanghai (1.414) 2.101 0.387 (0.572) (0.051) (0.281) (0.171) (1.414)

Spatial Weight Matrix multiplied by Z-Score Matrix (cell by cell multiplication)

Anhui Zhejiang Jiangxi Jiangsu Henan Hubei
wijzj
Shanghai SumWijZj LISA Lisa from
Zi 0.000 GeoDA
Anhui 2.101 - 0.077 (0.114) (0.010) (0.056) (0.034) - (0.137) -0.289 -0.248
Zhejiang 0.387 0.525 - (0.143) (0.013) - - (0.353) 0.016 0.006 0.005
Jiangxi (0.572) 0.700 0.129 - - - (0.057) - 0.772 -0.442 -0.379
Jiangsu (0.051) 0.700 0.129 - - - - (0.471) 0.358 -0.018 -0.016
Henan (0.281) 1.050 - - - - (0.085) - 0.965 -0.271 -0.233
Hubei (0.171) 0.700 - (0.191) - (0.094) - - 0.416 -0.071 -0.061
Shanghai (1.414) - 0.194 - (0.025) - - - 0.168 -0.238 -0.204

Source: Ron Briggs of UT Dallas 60

Local Getis-Ord G and G* Statistics
• Local Getis-Ord G
• It is the proportion of all x values in the study area
accounted for by the neighbors of location I
• G* will include the self value

w x ij j

G (d ) 
j

x
i
j
j

G will be high where high values cluster

G will be low where low values cluster  w (d )
ij

E (Gi (d )) 
j
Interpreted relative to expected value
if randomly distributed. n 1
61
 Moran Scatter Plot for GDI vs AL
Bivariate LISA
• Moran’s I is the correlation between X
and Lag-X--the same variable but in
nearby areas
– Univariate Moran’s I
• Bivariate Moran’s I is a correlation
between X and a different variable in
nearby areas.
Moran Significance Map for GDI vs. AL

62
Bivariate LISA
and the Correlation Coefficient
• Correlation Coefficient is the
relationship between two
different variables in the same
area
• Bivariate LISA is a correlation
between two different
variables in an area and in
nearby areas.

63
Bivariate Moran Scatter Plot

High/High
Low/High positive SA
negative SA

Low/Low High/Low
positive SA negative SA
64
 Summary
• Spatial autocorrelation of areal data
• Spatial weight matrix
• Measures of spatial autocorrelation
• Global Measure
– Moran’s I/Geary’s C/General G and G*
• Local
• LISA: Moran’s I/Geary’s C/General G and G*
• Bivariate LISA
– Significance test
65
• End of this topic

66