Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial...
Transcript of Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial...
![Page 1: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/1.jpg)
Areal Unit Data
Regular or Irregular Grids or LatticesLarge Point-referenced Datasets
Is there spatial pattern?
Chapter 3: Basics of Areal Data Models – p. 1/18
![Page 2: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/2.jpg)
Areal Unit Data
Regular or Irregular Grids or LatticesLarge Point-referenced Datasets
Is there spatial pattern?
Do we want to smooth the data?
Chapter 3: Basics of Areal Data Models – p. 1/18
![Page 3: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/3.jpg)
Areal Unit Data
Regular or Irregular Grids or LatticesLarge Point-referenced Datasets
Is there spatial pattern?
Do we want to smooth the data?
Inference for new areal units?
Chapter 3: Basics of Areal Data Models – p. 1/18
![Page 4: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/4.jpg)
Areal Unit Data
Regular or Irregular Grids or LatticesLarge Point-referenced Datasets
Is there spatial pattern?
Do we want to smooth the data?
Inference for new areal units?
Descriptive/algorithmic vs. Model-based
Chapter 3: Basics of Areal Data Models – p. 1/18
![Page 5: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/5.jpg)
Areal unit dataActual Transformed SIDS Rates
<2.02.0-3.03.0-3.5>3.5
Chapter 3: Basics of Areal Data Models – p. 2/18
![Page 6: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/6.jpg)
Proximity matrices
W , entries wij (with wii = 0). Choices for wij:
wij = 1 if i, j share a common boundary (possibly acommon vertex)wij is an inverse distance between unitswij = 1 if distance between units is ≤ K
wij = 1 for m nearest neighbors.
Chapter 3: Basics of Areal Data Models – p. 3/18
![Page 7: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/7.jpg)
Proximity matrices
W , entries wij (with wii = 0). Choices for wij:
wij = 1 if i, j share a common boundary (possibly acommon vertex)wij is an inverse distance between unitswij = 1 if distance between units is ≤ K
wij = 1 for m nearest neighbors.
W is typically symmetric, but need not be
Chapter 3: Basics of Areal Data Models – p. 3/18
![Page 8: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/8.jpg)
Proximity matrices
W , entries wij (with wii = 0). Choices for wij:
wij = 1 if i, j share a common boundary (possibly acommon vertex)wij is an inverse distance between unitswij = 1 if distance between units is ≤ K
wij = 1 for m nearest neighbors.
W is typically symmetric, but need not be
W : standardize row i by wi+ =∑
j wij (so matrix is nowrow stochastic, but probably no longer symmetric).
Chapter 3: Basics of Areal Data Models – p. 3/18
![Page 9: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/9.jpg)
Proximity matrices
W , entries wij (with wii = 0). Choices for wij:
wij = 1 if i, j share a common boundary (possibly acommon vertex)wij is an inverse distance between unitswij = 1 if distance between units is ≤ K
wij = 1 for m nearest neighbors.
W is typically symmetric, but need not be
W : standardize row i by wi+ =∑
j wij (so matrix is nowrow stochastic, but probably no longer symmetric).
W elements often called “weights”; interpretation
Chapter 3: Basics of Areal Data Models – p. 3/18
![Page 10: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/10.jpg)
Proximity matrices
W , entries wij (with wii = 0). Choices for wij:
wij = 1 if i, j share a common boundary (possibly acommon vertex)wij is an inverse distance between unitswij = 1 if distance between units is ≤ K
wij = 1 for m nearest neighbors.
W is typically symmetric, but need not be
W : standardize row i by wi+ =∑
j wij (so matrix is nowrow stochastic, but probably no longer symmetric).
W elements often called “weights”; interpretation
Could also define first-order neighbors W (1),second-order neighbors W (2), etc.
Chapter 3: Basics of Areal Data Models – p. 3/18
![Page 11: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/11.jpg)
Measures of spatial association
Moran’s I: essentially an “areal covariogram"
I =n
∑i
∑j wij(Yi − Y )(Yj − Y )
(∑
i 6=j wij)∑
i(Yi − Y )2
Chapter 3: Basics of Areal Data Models – p. 4/18
![Page 12: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/12.jpg)
Measures of spatial association
Moran’s I: essentially an “areal covariogram"
I =n
∑i
∑j wij(Yi − Y )(Yj − Y )
(∑
i 6=j wij)∑
i(Yi − Y )2
Geary’s C: essentially an “areal variogram"
C =(n − 1)
∑i
∑j wij(Yi − Yj)
2
(∑
i 6=j wij)∑
i(Yi − Y )2
Chapter 3: Basics of Areal Data Models – p. 4/18
![Page 13: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/13.jpg)
Measures of spatial association
Moran’s I: essentially an “areal covariogram"
I =n
∑i
∑j wij(Yi − Y )(Yj − Y )
(∑
i 6=j wij)∑
i(Yi − Y )2
Geary’s C: essentially an “areal variogram"
C =(n − 1)
∑i
∑j wij(Yi − Yj)
2
(∑
i 6=j wij)∑
i(Yi − Y )2
Both are asymptotically normal if Yi are i.i.d.;Moran has mean −1/(n − 1) ≈ 0, Geary has mean 1
Chapter 3: Basics of Areal Data Models – p. 4/18
![Page 14: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/14.jpg)
Measures of spatial association
Moran’s I: essentially an “areal covariogram"
I =n
∑i
∑j wij(Yi − Y )(Yj − Y )
(∑
i 6=j wij)∑
i(Yi − Y )2
Geary’s C: essentially an “areal variogram"
C =(n − 1)
∑i
∑j wij(Yi − Yj)
2
(∑
i 6=j wij)∑
i(Yi − Y )2
Both are asymptotically normal if Yi are i.i.d.;Moran has mean −1/(n − 1) ≈ 0, Geary has mean 1
Significance testing by comparing to a collection of say1000 random permutations of the Yi
Chapter 3: Basics of Areal Data Models – p. 4/18
![Page 15: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/15.jpg)
Measures of spatial association (cont’d)
<503504-525526-562>563
Figure 1: Choropleth map of 1999 average verbal
SAT scores, lower 48 U.S. states.Chapter 3: Basics of Areal Data Models – p. 5/18
![Page 16: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/16.jpg)
Measures of spatial association (cont’d)
For these data, the spatial.cor function inS+SpatialStats gives a Moran’s I of 0.5833, withassociated standard error estimate 0.0920 ⇒ verystrong evidence against H0 : no spatial correlation
Chapter 3: Basics of Areal Data Models – p. 6/18
![Page 17: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/17.jpg)
Measures of spatial association (cont’d)
For these data, the spatial.cor function inS+SpatialStats gives a Moran’s I of 0.5833, withassociated standard error estimate 0.0920 ⇒ verystrong evidence against H0 : no spatial correlation
spatial.cor also gives a Geary’s C of 0.3775, withassociated standard error estimate 0.1008 ⇒ again,very strong evidence against H0 (departure from 1)
Chapter 3: Basics of Areal Data Models – p. 6/18
![Page 18: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/18.jpg)
Measures of spatial association (cont’d)
For these data, the spatial.cor function inS+SpatialStats gives a Moran’s I of 0.5833, withassociated standard error estimate 0.0920 ⇒ verystrong evidence against H0 : no spatial correlation
spatial.cor also gives a Geary’s C of 0.3775, withassociated standard error estimate 0.1008 ⇒ again,very strong evidence against H0 (departure from 1)
Warning: These data have not been adjusted forcovariates, such as the proportion of students who takethe exam (Midwestern colleges have historically reliedon the ACT, not the SAT; only the best and brighteststudents in these states would bother taking the SAT)
Chapter 3: Basics of Areal Data Models – p. 6/18
![Page 19: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/19.jpg)
Measures of spatial association (cont’d)
For these data, the spatial.cor function inS+SpatialStats gives a Moran’s I of 0.5833, withassociated standard error estimate 0.0920 ⇒ verystrong evidence against H0 : no spatial correlation
spatial.cor also gives a Geary’s C of 0.3775, withassociated standard error estimate 0.1008 ⇒ again,very strong evidence against H0 (departure from 1)
Warning: These data have not been adjusted forcovariates, such as the proportion of students who takethe exam (Midwestern colleges have historically reliedon the ACT, not the SAT; only the best and brighteststudents in these states would bother taking the SAT)
⇒ the map, I, and C all motivate the search for spatialcovariates!
Chapter 3: Basics of Areal Data Models – p. 6/18
![Page 20: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/20.jpg)
Correlogram (via Moran’s I)
distance
rho(
d)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.0
-0.5
0.0
0.5
1.0
Replace wij with w(1)ij taken from W (1) ⇒ I(1)
Chapter 3: Basics of Areal Data Models – p. 7/18
![Page 21: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/21.jpg)
Correlogram (via Moran’s I)
distance
rho(
d)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.0
-0.5
0.0
0.5
1.0
Replace wij with w(1)ij taken from W (1) ⇒ I(1)
Replace wij with w(2)ij taken from W (2) ⇒ I(2), etc.
Chapter 3: Basics of Areal Data Models – p. 7/18
![Page 22: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/22.jpg)
Correlogram (via Moran’s I)
distance
rho(
d)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.0
-0.5
0.0
0.5
1.0
Replace wij with w(1)ij taken from W (1) ⇒ I(1)
Replace wij with w(2)ij taken from W (2) ⇒ I(2), etc.
Plot I(r) vs. r; an initial decline in r followed by variationaround 0 ⇒ spatial pattern!
Chapter 3: Basics of Areal Data Models – p. 7/18
![Page 23: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/23.jpg)
Correlogram (via Moran’s I)
distance
rho(
d)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.0
-0.5
0.0
0.5
1.0
Replace wij with w(1)ij taken from W (1) ⇒ I(1)
Replace wij with w(2)ij taken from W (2) ⇒ I(2), etc.
Plot I(r) vs. r; an initial decline in r followed by variationaround 0 ⇒ spatial pattern!
spatial analogue of the temporal lag autocorrelation plotChapter 3: Basics of Areal Data Models – p. 7/18
![Page 24: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/24.jpg)
Rasterized binary data map
NORTH SOUTH
land use classificationnon-forestforest
Chapter 3: Basics of Areal Data Models – p. 8/18
![Page 25: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/25.jpg)
Binary data correlogram
distance
logo
dds
0 50 100 150 200
01
23
N
0 20 40 60
01
23
NW
0 20 40 60
01
23
NE
0 10 20 30 40 50
01
23
E
A version for a binary map, using two-way tables andlog odds ratios at the pixel level
Chapter 3: Basics of Areal Data Models – p. 9/18
![Page 26: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/26.jpg)
Binary data correlogram
distance
logo
dds
0 50 100 150 200
01
23
N
0 20 40 60
01
23
NW
0 20 40 60
01
23
NE
0 10 20 30 40 50
01
23
E
A version for a binary map, using two-way tables andlog odds ratios at the pixel level
Note strongest pattern is to the north (N), but in nodirection are the values ≈ 0 even at 40 km
Chapter 3: Basics of Areal Data Models – p. 9/18
![Page 27: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/27.jpg)
Spatial smoothers
To smooth Yi, replace with Yi =∑
iwijYj
wi+
Chapter 3: Basics of Areal Data Models – p. 10/18
![Page 28: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/28.jpg)
Spatial smoothers
To smooth Yi, replace with Yi =∑
iwijYj
wi+
More generally, we could include the value actuallyobserved for unit i, and revise our smoother to
(1 − α)Yi + αYi
For 0 < α < 1, this is a linear (convex) combination in“shrinkage" form
Chapter 3: Basics of Areal Data Models – p. 10/18
![Page 29: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/29.jpg)
Spatial smoothers
To smooth Yi, replace with Yi =∑
iwijYj
wi+
More generally, we could include the value actuallyobserved for unit i, and revise our smoother to
(1 − α)Yi + αYi
For 0 < α < 1, this is a linear (convex) combination in“shrinkage" form
Finally, we could try model-based smoothing, i.e.,based on E(Yi|Data), i.e., the mean of the predictivedistribution. Smoothers then emerge as byproducts ofthe hierarchical spatial models we use to explain the Yi’s
Chapter 3: Basics of Areal Data Models – p. 10/18
![Page 30: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/30.jpg)
Markov random fields
Consider Y = (Y1, Y2, ..., Yn) and the set {p(yi|yj , j 6= i)}
Chapter 3: Basics of Areal Data Models – p. 11/18
![Page 31: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/31.jpg)
Markov random fields
Consider Y = (Y1, Y2, ..., Yn) and the set {p(yi|yj , j 6= i)}
We know p(y1, y2, ...yn) determines {p(yi|yj , j 6= i)}
(the set of full conditional distributions)
Chapter 3: Basics of Areal Data Models – p. 11/18
![Page 32: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/32.jpg)
Markov random fields
Consider Y = (Y1, Y2, ..., Yn) and the set {p(yi|yj , j 6= i)}
We know p(y1, y2, ...yn) determines {p(yi|yj , j 6= i)}
(the set of full conditional distributions)
Does {p(yi|yj , j 6= i)} determine p(y1, y2, ...yn)???
Chapter 3: Basics of Areal Data Models – p. 11/18
![Page 33: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/33.jpg)
Markov random fields
Consider Y = (Y1, Y2, ..., Yn) and the set {p(yi|yj , j 6= i)}
We know p(y1, y2, ...yn) determines {p(yi|yj , j 6= i)}
(the set of full conditional distributions)
Does {p(yi|yj , j 6= i)} determine p(y1, y2, ...yn)???
If the full conditionals are compatible, then by Brook’sLemma,
p(y1, . . . , yn) =p(y1|y2, . . . , yn)
p(y10|y2, . . . , yn)
p(y2|y10, y3, . . . , yn)
p(y20|y10, y3, . . . , yn)
. . .p(yn|y10, . . . , yn−1,0)
p(yn0|y10, . . . , yn−1,0)p(y10, . . . , yn0)
If left side is proper, the fact that it integrates to 1determines the normalizing constant!
Chapter 3: Basics of Areal Data Models – p. 11/18
![Page 34: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/34.jpg)
“Local” modeling
Suppose we specify the full conditionals such that
p(yi|yj , j 6= i) = p(yi|yj ∈ ∂i) ,
where ∂i is the set of neighbors of cell (region) i.When does {p(yi|yj ∈ ∂i)} determine p(y1, y2, ...yn)?
Chapter 3: Basics of Areal Data Models – p. 12/18
![Page 35: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/35.jpg)
“Local” modeling
Suppose we specify the full conditionals such that
p(yi|yj , j 6= i) = p(yi|yj ∈ ∂i) ,
where ∂i is the set of neighbors of cell (region) i.When does {p(yi|yj ∈ ∂i)} determine p(y1, y2, ...yn)?
Def’n: a clique is a set of cells such that each elementis a neighbor of every other element
Chapter 3: Basics of Areal Data Models – p. 12/18
![Page 36: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/36.jpg)
“Local” modeling
Suppose we specify the full conditionals such that
p(yi|yj , j 6= i) = p(yi|yj ∈ ∂i) ,
where ∂i is the set of neighbors of cell (region) i.When does {p(yi|yj ∈ ∂i)} determine p(y1, y2, ...yn)?
Def’n: a clique is a set of cells such that each elementis a neighbor of every other element
Def’n: a potential function of order k is a function of karguments that is exchangeable in these arguments
Chapter 3: Basics of Areal Data Models – p. 12/18
![Page 37: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/37.jpg)
“Local” modeling
Suppose we specify the full conditionals such that
p(yi|yj , j 6= i) = p(yi|yj ∈ ∂i) ,
where ∂i is the set of neighbors of cell (region) i.When does {p(yi|yj ∈ ∂i)} determine p(y1, y2, ...yn)?
Def’n: a clique is a set of cells such that each elementis a neighbor of every other element
Def’n: a potential function of order k is a function of karguments that is exchangeable in these arguments
Def’n: p(y1, . . . , yn) is a Gibbs distribution if, as afunction of the yi, it is a product of potentials on cliques
Chapter 3: Basics of Areal Data Models – p. 12/18
![Page 38: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/38.jpg)
“Local” modeling
Suppose we specify the full conditionals such that
p(yi|yj , j 6= i) = p(yi|yj ∈ ∂i) ,
where ∂i is the set of neighbors of cell (region) i.When does {p(yi|yj ∈ ∂i)} determine p(y1, y2, ...yn)?
Def’n: a clique is a set of cells such that each elementis a neighbor of every other element
Def’n: a potential function of order k is a function of karguments that is exchangeable in these arguments
Def’n: p(y1, . . . , yn) is a Gibbs distribution if, as afunction of the yi, it is a product of potentials on cliques
Example: For binary data, k = 2, we might takeQ(yi, yj) = I(yi = yj) = yiyj + (1 − yi)(1 − yj)
Chapter 3: Basics of Areal Data Models – p. 12/18
![Page 39: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/39.jpg)
“Local” modeling
Cliques of size 1 ⇔ independence
Chapter 3: Basics of Areal Data Models – p. 13/18
![Page 40: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/40.jpg)
“Local” modeling
Cliques of size 1 ⇔ independence
Cliques of size 2 ⇔ pairwise difference form
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i,j
(yi − yj)2I(i ∼ j)
and therefore p(yi|yj , j 6= i) = N(∑
j∈∂iyj/mi, τ
2/mi),where mi is the number of neighbors of i
Chapter 3: Basics of Areal Data Models – p. 13/18
![Page 41: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/41.jpg)
“Local” modeling
Cliques of size 1 ⇔ independence
Cliques of size 2 ⇔ pairwise difference form
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i,j
(yi − yj)2I(i ∼ j)
and therefore p(yi|yj , j 6= i) = N(∑
j∈∂iyj/mi, τ
2/mi),where mi is the number of neighbors of i
Hammersley-Clifford Theorem: If we have a MarkovRandom Field (i.e., {p(yi|yj ∈ ∂i)} uniquely determinep(y1, y2, ...yn)), then the latter is a Gibbs distribution
Chapter 3: Basics of Areal Data Models – p. 13/18
![Page 42: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/42.jpg)
“Local” modeling
Cliques of size 1 ⇔ independence
Cliques of size 2 ⇔ pairwise difference form
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i,j
(yi − yj)2I(i ∼ j)
and therefore p(yi|yj , j 6= i) = N(∑
j∈∂iyj/mi, τ
2/mi),where mi is the number of neighbors of i
Hammersley-Clifford Theorem: If we have a MarkovRandom Field (i.e., {p(yi|yj ∈ ∂i)} uniquely determinep(y1, y2, ...yn)), then the latter is a Gibbs distribution
Geman and Geman result : If we have a joint Gibbsdistribution, then we have a Markov Random Field
Chapter 3: Basics of Areal Data Models – p. 13/18
![Page 43: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/43.jpg)
Conditional autoregressive (CAR) modelGaussian (autonormal) case
p(yi|yj , j 6= i) = N
∑
j
bijyj , τ2i
Using Brook’s Lemma we can obtain
p(y1, y2, ...yn) ∝ exp
{−
1
2y′D−1(I − B)y
}
where B = {bij} and D is diagonal with Dii = τ2i .
Chapter 3: Basics of Areal Data Models – p. 14/18
![Page 44: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/44.jpg)
Conditional autoregressive (CAR) modelGaussian (autonormal) case
p(yi|yj , j 6= i) = N
∑
j
bijyj , τ2i
Using Brook’s Lemma we can obtain
p(y1, y2, ...yn) ∝ exp
{−
1
2y′D−1(I − B)y
}
where B = {bij} and D is diagonal with Dii = τ2i .
⇒ suggests a multivariate normal distribution withµY = 0 and ΣY = (I − B)−1D
Chapter 3: Basics of Areal Data Models – p. 14/18
![Page 45: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/45.jpg)
Conditional autoregressive (CAR) modelGaussian (autonormal) case
p(yi|yj , j 6= i) = N
∑
j
bijyj , τ2i
Using Brook’s Lemma we can obtain
p(y1, y2, ...yn) ∝ exp
{−
1
2y′D−1(I − B)y
}
where B = {bij} and D is diagonal with Dii = τ2i .
⇒ suggests a multivariate normal distribution withµY = 0 and ΣY = (I − B)−1D
D−1(I − B) symmetric requires bij
τ2i
= bji
τ2j
for all i, j
Chapter 3: Basics of Areal Data Models – p. 14/18
![Page 46: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/46.jpg)
CAR Model (cont’d)Returning to W , let bij = wij/wi+ and τ2
i = τ2/wi+, so
p(y1, y2, ...yn) ∝ exp
{−
1
2τ2y′(Dw − W )y
}
where Dw is diagonal with (Dw)ii = wi+ and thus
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i 6=j
wij(yi − yj)2
Intrinsic autoregressive (IAR) model!
Chapter 3: Basics of Areal Data Models – p. 15/18
![Page 47: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/47.jpg)
CAR Model (cont’d)Returning to W , let bij = wij/wi+ and τ2
i = τ2/wi+, so
p(y1, y2, ...yn) ∝ exp
{−
1
2τ2y′(Dw − W )y
}
where Dw is diagonal with (Dw)ii = wi+ and thus
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i 6=j
wij(yi − yj)2
Intrinsic autoregressive (IAR) model!
Joint distribution is improper since (Dw − W )1 = 0, sorequires a constraint – say,
∑i yi = 0
Chapter 3: Basics of Areal Data Models – p. 15/18
![Page 48: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/48.jpg)
CAR Model (cont’d)Returning to W , let bij = wij/wi+ and τ2
i = τ2/wi+, so
p(y1, y2, ...yn) ∝ exp
{−
1
2τ2y′(Dw − W )y
}
where Dw is diagonal with (Dw)ii = wi+ and thus
p(y1, y2, ...yn) ∝ exp
−
1
2τ2
∑
i 6=j
wij(yi − yj)2
Intrinsic autoregressive (IAR) model!
Joint distribution is improper since (Dw − W )1 = 0, sorequires a constraint – say,
∑i yi = 0
Not a data model – a random effects model!
Chapter 3: Basics of Areal Data Models – p. 15/18
![Page 49: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/49.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 50: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/50.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 51: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/51.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
“To ρ or not to ρ?” (that is the question)
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 52: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/52.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
“To ρ or not to ρ?” (that is the question)
Advantages:
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 53: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/53.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
“To ρ or not to ρ?” (that is the question)
Advantages:
makes distribution proper
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 54: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/54.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
“To ρ or not to ρ?” (that is the question)
Advantages:
makes distribution proper
adds parametric flexibility
Chapter 3: Basics of Areal Data Models – p. 16/18
![Page 55: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/55.jpg)
CAR Model IssuesWith τ2 unknown, appropriate exponent for τ2?Hodges et al.: n − I, where I is the # of “islands”
“Proper version:” replace Dw − W by Dw − ρW , andchoose ρ so that Σy = (Dw − ρW )−1 exists!This in turn implies Yi|Yj 6=i ∼ N(ρ
∑j wijYj , τ
2/mi)
“To ρ or not to ρ?” (that is the question)
Advantages:
makes distribution proper
adds parametric flexibility
ρ = 0 interpretable as independenceChapter 3: Basics of Areal Data Models – p. 16/18
![Page 56: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/56.jpg)
CAR Models with ρ parameter
Disadvantages:
why should we expect Yi to be a proportion of averageof neighbors – a sensible spatial interpretation?
Chapter 3: Basics of Areal Data Models – p. 17/18
![Page 57: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/57.jpg)
CAR Models with ρ parameter
Disadvantages:
why should we expect Yi to be a proportion of averageof neighbors – a sensible spatial interpretation?
calibration of ρ as a correlation, e.g.,
ρ = 0.80 yields 0.1 ≤ Moran’s I ≤ 0.15,
ρ = 0.90 yields 0.2 ≤ Moran’s I ≤ 0.25,
ρ = 0.99 yields Moran’s I ≤ 0.5
Chapter 3: Basics of Areal Data Models – p. 17/18
![Page 58: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/58.jpg)
CAR Models with ρ parameter
Disadvantages:
why should we expect Yi to be a proportion of averageof neighbors – a sensible spatial interpretation?
calibration of ρ as a correlation, e.g.,
ρ = 0.80 yields 0.1 ≤ Moran’s I ≤ 0.15,
ρ = 0.90 yields 0.2 ≤ Moran’s I ≤ 0.25,
ρ = 0.99 yields Moran’s I ≤ 0.5
So, used with random effects, scope of spatial patternmay be limited
Chapter 3: Basics of Areal Data Models – p. 17/18
![Page 59: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/59.jpg)
Comments
The CAR specifies Σ−1Y , not ΣY (as in the point-level
modeling case), so does not directly model association
Chapter 3: Basics of Areal Data Models – p. 18/18
![Page 60: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/60.jpg)
Comments
The CAR specifies Σ−1Y , not ΣY (as in the point-level
modeling case), so does not directly model association
(Σ−1Y )ii = 1/τ2
i ; (Σ−1Y )ij = 0 ⇔ cond’l independence
Chapter 3: Basics of Areal Data Models – p. 18/18
![Page 61: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/61.jpg)
Comments
The CAR specifies Σ−1Y , not ΣY (as in the point-level
modeling case), so does not directly model association
(Σ−1Y )ii = 1/τ2
i ; (Σ−1Y )ij = 0 ⇔ cond’l independence
Prediction at new sites is ad hoc, in that if
p(y0|y1, y2, ...yn) = N(∑
j
w0jyj/w0+, τ2/w0+)
then p(y0, y1, ...yn) well-defined but not CAR
Chapter 3: Basics of Areal Data Models – p. 18/18
![Page 62: Is there spatial pattern? - School of Public Healthbrad/8472/slide3_2day.pdf · Is there spatial pattern? Do we want to smooth the data? ... wij = 1 if i,j share a common boundary](https://reader031.fdocuments.us/reader031/viewer/2022022506/5ac3bad47f8b9aae1b8cbc45/html5/thumbnails/62.jpg)
Comments
The CAR specifies Σ−1Y , not ΣY (as in the point-level
modeling case), so does not directly model association
(Σ−1Y )ii = 1/τ2
i ; (Σ−1Y )ij = 0 ⇔ cond’l independence
Prediction at new sites is ad hoc, in that if
p(y0|y1, y2, ...yn) = N(∑
j
w0jyj/w0+, τ2/w0+)
then p(y0, y1, ...yn) well-defined but not CAR
Non-Gaussian case: For binary data, the autologistic:
p(yi|yj , j 6= i) ∝ exp
φ
∑
j
wijI(yi = yj)
Chapter 3: Basics of Areal Data Models – p. 18/18