1 spatial processes and statistical modelling peter green university of bristol, uk ims/isba, san...
TRANSCRIPT
![Page 1: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/1.jpg)
1
Spatial processes and statistical modelling
Peter GreenUniversity of Bristol, UK
IMS/ISBA, San Juan, 24 July 2003
![Page 2: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/2.jpg)
2
• Continuous space
• Discrete space– lattice– irregular - general graphs– areally aggregated
• Point processes– other object processes
Spatial indexing
![Page 3: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/3.jpg)
3
Purpose of overview
• setting the scene for 8 invited talks on spatial statistics
• particularly for specialists in the other 2 areas
![Page 4: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/4.jpg)
4
Perspective of overview
• someone interested in the development of methodology– for the analysis of spatially-indexed
data– probably Bayesian
• models and frameworks, not applications
• personal, selective, eclectic
![Page 5: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/5.jpg)
5
Genesis of spatial statistics• adaptation of time series ideas• ‘applied probability’ modelling• geostatistics• application-led
![Page 6: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/6.jpg)
6
Space vs. time
• apparently slight difference• profound implications for
mathematical formulation and computational tractability
![Page 7: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/7.jpg)
7
Requirements of particular application domains• agriculture (design)
• ecology (sparse point pattern, poor data?)
• environmetrics (space/time)
• climatology (huge physical models)
• epidemiology (multiple indexing)
• image analysis (huge size)
![Page 8: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/8.jpg)
8
Key themes
• conditional independence– graphical/hierarchical modelling
• aggregation– analysing dependence between differently
indexed data– opportunities and obstacles
• literal credibility of models• Bayes/non-Bayes distinction blurred
![Page 9: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/9.jpg)
9
A big subject….
Noel Cressie: “This may be the last time spatial
statistics is squeezed between two covers”
(Preface to Statistics for Spatial Data, 900pp., Wiley, 1991)
![Page 10: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/10.jpg)
10
Why build spatial dependence into a model?• No more reason to suppose
independence in spatially-indexed data than in a time-series
• However, substantive basis for form of spatial dependent sometimes slight - very often space is a surrogate for missing covariates that are correlated with location
![Page 11: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/11.jpg)
11
Discretely indexed data
![Page 12: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/12.jpg)
12
Modelling spatial dependence in discretely-indexed fields• Direct• Indirect
– Hidden Markov models– Hierarchical models
![Page 13: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/13.jpg)
13
Hierarchical models, using DAGs
Variables at several levels - allows modelling of complex systems, borrowing strength, etc.
![Page 14: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/14.jpg)
14
Modelling with undirected graphsDirected acyclic graphs are a natural
representation of the way we usually specify a statistical model - directionally:
• disease symptom• past future• parameters data ……whether or not causality is understood.But sometimes (e.g. spatial models) there is
no natural direction
![Page 15: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/15.jpg)
15
Conditional independence
In model specification, spatial context often rules out directional dependence (that would have been acceptable in time series context)
X0 X1 X2 X3 X4
![Page 16: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/16.jpg)
16
Conditional independence
In model specification, spatial context often rules out directional dependence
X20 X21 X22 X23 X24
X00 X01 X02 X03 X04
X10 X11 X12 X13 X14
![Page 17: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/17.jpg)
17
Conditional independence
In model specification, spatial context often rules out directional dependence
X20 X21 X22 X23 X24
X00 X01 X02 X03 X04
X10 X11 X12 X13 X14
![Page 18: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/18.jpg)
18
Directed acyclic graph
)|()( )(pa vVv
v xxpxp
in general:
for example:
a b
c
dp(a,b,c,d)=p(a)p(b)p(c|a,b)p(d|c)In the RHS, any distributions are legal, and uniquely define joint distribution
![Page 19: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/19.jpg)
19
Undirected (CI) graph
X20 X21 X22
X00 X01 X02
X10 X11 X12Absence of edge denotes conditional independence given all other variables
But now there are non-trivial constraints on conditional distributions
Regular lattice, irregular graph, areal data...
![Page 20: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/20.jpg)
20
Undirected (CI) graph
X20 X21 X22
X00 X01 X02
X10 X11 X12
C
CC XVXp ))(exp)(
iC
CCii XVXXp ))(exp)|(
)|()|( iiii XXpXXp
The Hammersley-Clifford theorem says essentially that the converse is also true - the only sure way to get a valid joint distribution is to use ()
()
clique
![Page 21: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/21.jpg)
21
Hammersley-Clifford
X20 X21 X22
X00 X01 X02
X10 X11 X12
C
CC XVXp )(exp)(
)|()|( iiii XXpXXp
A positive distribution p(X) is a Markov random field
if and only if it is a Gibbs distribution
- Sum over cliques C (complete subgraphs)
![Page 22: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/22.jpg)
22
Partition function
X20 X21 X22
X00 X01 X02
X10 X11 X12
C
CC XVXp )(exp)(
Almost always, the constant of proportionality in
is not available in tractable form: an obstacle to likelihood or Bayesian inference about parameters in the potential functions
Physicists call
the partition function
))( CC XV
X C
CC XVZ )(exp
![Page 23: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/23.jpg)
23
Gaussian Markov random fields: spatial autoregression
C
CC XVXp )(exp)(
)|()|( iiii XXpXXp is a multivariate Gaussian distribution, and
If VC(XC) is -ij(xi-xj)2/2 for C={i,j} and 0 otherwise, then
is the univariate Gaussian distribution
),(~| 22i
ijjijiii XNXX
ij
iji /12where
![Page 24: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/24.jpg)
24
Inverse of (co)variance matrix:
dependent case3210
2410
1121
0012
A B C D
non-zero
non-zero),|,cov( CADB
A B C D
A
B
C
D
Gaussian random fields
![Page 25: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/25.jpg)
25
Gaussian Markov random fields: spatial autoregressionDistinguish these conditional autoregression (CAR) models from the corresponding simultaneous autoregression (SAR) models
j
ijiji XX ~
(cf time series case).
The latter are less compatible with hierarchical model structures.
i i.i.d. normal
![Page 26: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/26.jpg)
26
Non-Gaussian Markov random fields
ji
ji xx
~
)cosh(log)1(exp
C
CC XVXp )(exp)(
Pairwise interaction random fields with less smooth realisations obtained by replacing squared differences by a term with smaller tails, e.g.
![Page 27: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/27.jpg)
27
Agricultural field trials
• strong cultural constraints• design, randomisation, cultivation effects• 1-D analysis in 2-d fields• relationships between IB designs, splines,
covariance models, spatial autoregression…
![Page 28: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/28.jpg)
28
Discrete Markov random fields
C
CC XVXp )(exp)(
Besag (1974) introduced various cases of
for discrete variables, e.g. auto-logistic (binary variables), auto-Poisson (local conditionals are Poisson), auto-binomial, etc.
![Page 29: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/29.jpg)
29
Auto-logistic model
C
CC XVXp )(exp)(
ji
jiiji
ii xxx~
exp
)|()|( iiii XXpXXp is Bernoulli(pi) with
ij
jijiii xpp )())1/(log(
- a very useful model for dependent binary variables (NB various parameterisations)
(Xi = 0 or 1)
![Page 30: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/30.jpg)
30
Statistical mechanics models
C
CC XVXp )(exp)(
The classic Ising model (for ferromagnetism) is the symmetric autologistic model on a square lattice in 2-D or 3-D. The Potts model is the generalisation to more than 2 ‘colours’
ji
jii
x xxIXpi
~
][exp)(
and of course you can usefully un-symmetrise this.
![Page 31: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/31.jpg)
31
Auto-Poisson model
C
CC XVXp )(exp)(
ji
jiiji
iii xxxx~
)!log(exp
)|()|( iiii XXpXXp is Poisson ))(exp(
ij
jiji x
For integrability, ij must be 0, so this onlymodels negative dependence: very limited use.
![Page 32: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/32.jpg)
32
Hierarchical models and
hidden Markov processes
![Page 33: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/33.jpg)
33
Chain graphs
• If both directed and undirected edges, but no directed loops:
• can rearrange to form global DAG with undirected edges within blocks
![Page 34: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/34.jpg)
34
Chain graphs
• If both directed and undirected edges, but no directed loops:
• can rearrange to form global DAG with undirected edges within blocks
• Hammersley-Clifford within blocks
![Page 35: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/35.jpg)
35
Hidden Markov random fields• We have a lot of freedom modelling
spatially-dependent continuously-distributed random fields on regular or irregular graphs
• But very little freedom with discretely distributed variables
use hidden random fields, continuous or discrete
• compatible with introducing covariates, etc.
![Page 36: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/36.jpg)
36
Hidden Markov models
z0 z1 z2 z3 z4
y1 y2 y3 y4
e.g. Hidden Markov chain
observed
hidden
![Page 37: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/37.jpg)
37
Hidden Markov random fields
Unobserved dependent field
Observed conditionally-independent discrete field
(a chain graph)
![Page 38: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/38.jpg)
38
Spatial epidemiology applications
independently, for each region i. Options:• CAR, CAR+white noise (BYM, 1989)• Direct modelling of ,e.g. SAR• Mixture/allocation/partition models:
• Covariates, e.g.:
)(~ iii ePoissonY casesexpected
cases
relative risk
)log( i))cov(log( i
izi eYEi
)(
)exp()( Tiizi xeYE
i
![Page 39: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/39.jpg)
39
Spatial epidemiology applications
Spatial contiguity is usually somewhat idealised
![Page 40: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/40.jpg)
40
CAR model for lip cancer data
observed counts
random spatial effects
covariate
regressioncoefficient
expected counts
(WinBUGS example)
![Page 41: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/41.jpg)
41
relativerisk
parameters
Example of an allocation modelRichardson & Green (JASA, 2002) used a
hidden Markov random field model for disease mapping
)(Poisson~ izi eyi
observedincidence
expectedincidencehidden
MRF
![Page 42: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/42.jpg)
42)(Poisson~ ii eY
Chain graph for disease mapping
e
Y
zk
)(Poisson~ izi eYi
based on Potts model
![Page 43: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/43.jpg)
43
Larynx cancer in females in France
SMRs
)|1( ypiz
ii Ey /
![Page 44: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/44.jpg)
44
Continuously indexed data
![Page 45: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/45.jpg)
45
Continuously indexed fields
The basic model is the Gaussian random field
with and
Translation-invariant or fully stationary (isotropic) cases have
and or ,resp.
2),( EssZ)())(( ssZE
),())(),(cov( tsctZsZ
))(( sZE )())(),(cov( tsctZsZ ||)(||))(),(cov( tsctZsZ
![Page 46: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/46.jpg)
46
Geostatistics and kriging
• There is a huge literature on a group of methodologies originally developed for geographical and geological data
• The main theme is prediction of (functionals of) a random field based on observations at a finite set of locations
![Page 47: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/47.jpg)
47
Ordinary kriging
• is a random process, we have observations and we wish to predict , e.g. a “block average”
• The usual basis is least-squares prediction, using a model for the mean and covariance of estimated from the data
2),( EssZ )(),...,(),( 21 nsZsZsZ
(.))(Zg
B
BdssZ ||/)(
)(sZ
![Page 48: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/48.jpg)
48
Ordinary kriging
The usual assumption is that is intrinsically stationary, i.e. has 2nd order structure
for all s is called the semi-variogramThis is somewhat weaker than full 2nd-
order stationarity
)(sZ
)(2))()(var( hsZhsZ
)(h
0))()(( sZhsZE
![Page 49: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/49.jpg)
49
Ordinary kriging
The optimal solution to the prediction problem in terms of the semivariogram follows from standard linear algebra arguments; an empirical estimate of the semivariogram is then plugged in.
![Page 50: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/50.jpg)
50
Variants of kriging
Kriging without intrinsic stationarity (& a model instead of empirical estimates)
Co-kriging (multivariate)Robust krigingUniversal kriging (kriging with regression)Disjunctive (nonlinear) krigingIndicator krigingConnections with splines
![Page 51: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/51.jpg)
51
Given data (si,xi,Yi), build model starting
with a Gaussian random fieldwith andSet whereand
Bayesian geostatistics (Diggle, Moyeed and Tawn, Appl Stat, 1998)
2),( EssZ0))(( sZE )())(),(cov( tsctZsZ
)|(~(.)| iii MyfZY (.))|( ZYEM ii T
iii xsZMh )()(
Z*
Z
Y
Y*
inference
prediction
X
X*
Z
![Page 52: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/52.jpg)
52
Point data
![Page 53: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/53.jpg)
53
Point processes
• (inhomogeneous) Poisson process• Neyman-Scott process• (log Gaussian) Cox process• Gibbs point process• Markov point process • Area-interaction process
![Page 54: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/54.jpg)
54
Analysis of spatial point pattern• Very strong early emphasis on
modelling clustering and repelling alternatives to homogeneous Poisson process (complete spatial randomness)
• May be different effects at different scales
• Interpretations in terms of mechanisms, e.g. in ecology, forestry
![Page 55: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/55.jpg)
55
Point process as parametrisation of spaceVoronoi tessellation of random point
process
Flexible modelling of surfaces: step functions, polynomials, …
![Page 56: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/56.jpg)
56
Rare disease point data• Regard locations of cases as Poisson
process with highly structured intensity process– Covariates– Spatial dependence
)()())(( dsesdsYE
As
dsesAYE )()())((
)(dsY number of cases in ds
![Page 57: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/57.jpg)
57
Models without covariates 1
Cox processwhere is a random field, e.g.
is Gaussian – ‘log Gaussian Cox process’ (Moller, Syversveen and Waagepetersen, 1998)
)(s)())(log( sZs
)()())(( dsesdsYE
![Page 58: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/58.jpg)
58
Models without covariates 2Smoothed Gamma random field
(Wolpert and Ickstadt, 1998)
where is a kernel function
and
is a sum of smoothed gamma-distributed impulses
-- example of shot-noise Cox process
)(),()( duusks
))(),((~)( ududu ),( usk
)(s
![Page 59: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/59.jpg)
59
DAG for Gamma RF model with covariates
k
Y
eX
measure
pointprocess
function
vector
key
![Page 60: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/60.jpg)
60
Models without covariates 3
Voronoi tessellation models(PJG,1995; Heikkinen and Arjas, 1998)
where are cells of Voronoi tessellation of an unobserved point process and
might be independent or dependent (e.g. CAR model for logs)
k
kk AsIs ][)( kA
k
![Page 61: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/61.jpg)
61
Introducing covariates
With covariates {Xj(s)} measured at case locations s, usual formulation is multiplicative
but occasionally additive
+ data-dependent constraints on parameters
j
jj sXss )(exp)()(
j
jj sXss )()()(
![Page 62: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/62.jpg)
62
Markov point processesRich families of non-Poisson point processes can be defined by
specifying their densities (Radon-Nikodym derivatives) w.r.t. unit-rate Poisson process, e.g. pairwise interaction models
(e.g. g(si,sj)=<1 if d(si,sj)<, 1 otherwise), and
area-interaction models
Note formal similarity to Gibbs lattice modelsMarginal distribution of #points usually not explicit
),(
)( ),()(ji
jisn ssgsp
||)()( Tssnsp +++
++
+
![Page 63: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/63.jpg)
63
Object processes
• Poisson processes of objects (lines, planes, flats, ….)
• Coloured triangulations….
![Page 64: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/64.jpg)
64
Aggregation
![Page 65: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/65.jpg)
65
Aggregation coherence and ecological bias• Commonly, covariates and responses are
spatially indexed differently, and for most models this poses coherence problems (linear Gaussian case the main exception)
• E.g. areally-aggregated response Yi=Y(Ai), and continuously indexed covariate X(s)
iAs
i dsesXsAYE )())(exp()())((
![Page 66: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/66.jpg)
66
Aggregation coherence and ecological bias
Even with uniform , this is not of form
where
(mis-specification) bias in estimation of .Need to know spatial variation in covariate
iAs
i dsesXsAYE )())(exp()())((
)()( dses
))(exp())(( iii AXAYE
iA
ii AdssXAX ||/)()(
![Page 67: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/67.jpg)
67
Aggregation coherence and ecological bias
Additive formulation
avoids this problem, as does the Ickstadt and Wolpert approach, to some extent
iAs
i dsesXsAYE )())()(())((
![Page 68: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/68.jpg)
68
Invited talks on spatial statistics• Brad Carlin: space & space-time CDF models, air
pollutant data• Jon Wakefield: ecological fallacy• Montserrat Fuentes: spatial design, air pollution• Doug Nychka: filtering for weather forecasting• Susie Bayarri: validating computer models• Arnoldo Frigessi: localisation of GSM phones• Rasmus Waagepetersen: Poisson-log Gaussian
processes• Adrian Baddeley: point process diagnostics
Fri 09:00
Sat 10:45
Fri 17:30
Fri 10:45
![Page 69: 1 Spatial processes and statistical modelling Peter Green University of Bristol, UK IMS/ISBA, San Juan, 24 July 2003](https://reader030.vdocuments.us/reader030/viewer/2022032605/56649e715503460f94b70002/html5/thumbnails/69.jpg)
69
Spatial processes and statistical modelling
[email protected]://www.stats.bris.ac.uk/~peter/PR
Peter GreenUniversity of Bristol, UKIMS/ISBA, San Juan, 24 July 2003