The Effect on Attribute Prediction of Location Uncertainty in Spatial Data

A datum is considered spatial if it contains location information. Typically, there is also attribute information, whose distribution depends on its location. Thus, error in location information can lead to error in attribute information, which is reflected ultimately in the inference drawn from the data. We propose a statistical model for incorporating location error into spatial data analysis. We investigate the effect of location error on the spatial lag, the covariance function, and optimal spatial linear prediction (that is, kriging). We show that the form of kriging after adjusting for location error is the same as that of kriging without adjusting for location error. However, location error changes entries in the matrix of explanatory variables, the matrix of co‐variances between the sample sites, and the vector of covariances between the sample sites and the prediction location. We investigate, through simulation, the effect that varying trend, measurement error, location error, range of spatial dependence, sample size, and prediction location have on kriging after and without adjusting for location error. When the location error is large, kriging after adjusting for location error performs markedly better than kriging without adjusting for location error, in terms of both the prediction bias and the mean squared prediction error.     

A Geostatistical Framework for Area‐to‐Point Spatial Interpolation

The spatial prediction of point values from areal data of the same attribute is addressed within the general geostatistical framework of change of support; the term support refers to the domain informed by each datum or unknown value. It is demonstrated that the proposed geostatistical framework can explicitly and consistently account for the support differences between the available areal data and the sought‐after point predictions. In particular, it is proved that appropriate modeling of all area‐to‐area and area‐to‐point covariances required by the geostatistical frame‐work yields coherent (mass‐preserving or pycnophylactic) predictions. In other words, the areal average (or areal total) of point predictions within any arbitrary area informed by an areal‐average (or areal‐total) datum is equal to that particular datum. In addition, the proposed geostatistical framework offers the unique advantage of providing a measure of the reliability (standard error) of each point prediction. It is also demonstrated that several existing approaches for area‐to‐point interpolation can be viewed within this geostatistical framework. More precisely, it is shown that (i) the choropleth map case corresponds to the geostatistical solution under the assumption of spatial independence at the point support level; (ii) several forms of kernel smoothing can be regarded as alternative (albeit sometimes incoherent) implementations of the geostatistical approach; and (iii) Tobler's smooth pycnophylactic interpolation, on a quasi‐infinite domain without non‐negativity constraints, corresponds to the geostatistical solution when the semivariogram model adopted at the point support level is identified to the free‐space Green's functions (linear in 1‐D or logarithmic in 2‐D) of Poisson's partial differential equation. In lieu of a formal case study, several 1‐D examples are given to illustrate pertinent concepts.     

A Simulation Study on Specifying a Regression Model for Spatial Data: Choosing between Autocorrelation and Heterogeneity Effects

In this simulation study, regressions specified with autocorrelation effects are compared against those with relationship heterogeneity effects, and in doing so, provides guidance on their use. Regressions investigated are: (1) multiple linear regression, (2) a simultaneous autoregressive error model, and (3) geographically weighted regression. The first is nonspatial and acts as a control, the second accounts for stationary spatial autocorrelation via the error term, while the third captures spatial heterogeneity through the modeling of nonstationary relationships between the response and predictor variables. The geostatistical‐based simulation experiment generates data and coefficients with known multivariate spatial properties, all within an area‐unit spatial setting. Spatial autocorrelation and spatial heterogeneity effects are varied and accounted for. On fitting the regressions, that each have different assumptions and objectives, to very different geographical processes, valuable insights to their likely performance are uncovered. Results objectively confirm an inherent interrelationship between autocorrelation and heterogeneity, that results in an identification problem when choosing one regression over another. Given this, recommendations on the use and implementation of these spatial regressions are suggested, where knowledge of the properties of real study data and the analytical questions being posed are paramount.     

A Geostatistical Linear Regression Model for Small Area Data. 一种适用于小区域数据的地统计线性回归模型

We present a new linear regression model for use with aggregated, small area data that are spatially autocorrelated. Because these data are aggregates of individual‐level data, we choose to model the spatial autocorrelation using a geostatistical model specified at the scale of the individual. The autocovariance of observed small area data is determined via the natural aggregation over the population. Unlike lattice‐based autoregressive approaches, the geostatistical approach is invariant to the scale of data aggregation. We establish that this geostatistical approach also is a valid autoregressive model; thus, we call this approach the geostatistical autoregressive (GAR) model. An asymptotically consistent and efficient maximum likelihood estimator is derived for the GAR model. Finite sample evidence from simulation experiments demonstrates the relative efficiency properties of the GAR model. Furthermore, while aggregation results in less efficient estimates than disaggregated data, the GAR model provides the most efficient estimates from the data that are available. These results suggest that the GAR model should be considered as part of a spatial analyst's toolbox when aggregated, small area data are analyzed. More important, we believe that the GAR model's attention to the individual‐level scale allows for a more flexible and theory‐informed specification than the existing autoregressive approaches based on an area‐level spatial weights matrix. Because many spatial process models, both in geography and in other disciplines, are specified at the individual level, we hope that the GAR covariance specification will provide a vehicle for a better informed and more interdisciplinary use of spatial regression models with area‐aggregated data.     

Adaptive Cell Tower Location Using Geostatistics. 基于地统计学的适应性基站发射塔选址研究

In this article, we address the problem of allocating an additional cell tower (or a set of towers) to an existing cellular network, maximizing the call completion probability. Our approach is derived from the adaptive spatial sampling problem using kriging, capitalizing on spatial correlation between cell phone signal strength data points and accounting for terrain morphology. Cell phone demand is reflected by population counts in the form of weights. The objective function, which is the weighted call completion probability, is highly nonlinear and complex (nondifferentiable and discontinuous). Sequential and simultaneous discrete optimization techniques are presented, and heuristics such as simulated annealing and Nelder–Mead are suggested to solve our problem. The adaptive spatial sampling problem is defined and related to the additional facility location problem. The approach is illustrated using data on cell phone call completion probability in a rural region of Erie County in western New York, and accounts for terrain variation using a line‐of‐sight approach. Finally, the computational results of sequential and simultaneous approaches are compared. Our model is also applicable to other facility location problems that aim to minimize the uncertainty associated with a customer visiting a new facility that has been added to an existing set of facilities.     

Mapping the possible occurrence of archaeological sites by Bayesian inference

Regional archaeological prospections are often done by field walking, where the location of the sampled fields is often determined by factors like feared disturbance or recent plowing. The resulting data configuration can be suboptimal for spatial prediction of the archaeological potential by geostatistical methods like kriging. As an alternative, we propose a Bayesian method to map the possible occurrence of archaeological finds and compare this to indicator regression kriging. Three types of predictive models were implemented in the Bayesian context following deductive, inductive and mixed approaches to use auxiliary geographical information in the mapping. After prediction of a validation set, it was concluded that the mixed approach gave the best results in terms of map quality, and that the kriging method performed poorly. Usage of data on the presence and the absence of archaeological finds is to be preferred above usage of presence data only. Furthermore, a method is presented that filters those parts of a predictive map that are not strongly supported by evidence.     

Comparison of Moran's I and Geary's c in Multivariate Spatial Pattern Analysis

This article compares multivariate spatial analysis methods that include not only multivariate covariance, but also spatial dependence of the data explicitly and simultaneously in model design by extending two univariate autocorrelation measures, namely Moran's I and Geary's c. The results derived from the simulation datasets indicate that the standard Moran component analysis is preferable to Geary component analysis as a tool for summarizing multivariate spatial structures. However, the generalized Geary principal component analysis developed in this study by adding variance into the optimization criterion and solved as a trace ratio optimization problem performs as well as, if not better than its counterpart the Moran principal component analysis does. With respect to the sensitivity in detecting subtle spatial structures, the choice of the appropriate tool is dependent on the correlation and variance of the spatial multivariate data. Finally, the four techniques are applied to the Social Determinants of Health dataset to analyze its multivariate spatial pattern. The two generalized methods detect more urban areas and higher autocorrelation structures than the other two standard methods, and provide more obvious contrast between urban and rural areas due to the large variance of the spatial component.     

Nonparametric Significance Test for Discovery of Network‐Constrained Spatial Co‐Location Patterns

Spatial co‐location patterns are useful for understanding positive spatial interactions among different geographical phenomena. Existing methods for detecting spatial co‐location patterns are mostly developed based on planar space assumption; however, geographical phenomena related to human activities are strongly constrained by road networks. Although these methods can be simply modified to consider the constraints of networks by using the network distance or network partitioning scheme, user‐specified parameters or priori assumptions for determining prevalent co‐location patterns are still subjective. As a result, some co‐location patterns may be wrongly reported or omitted. Therefore, a nonparametric significance test without priori assumptions about the distributions of the spatial features is proposed in this article. Both point‐dependent and location‐dependent network‐constrained summary statistics are first utilized to model the distribution characteristics of the spatial features. Then, by using these summary statistics, a network‐constrained pattern reconstruction method is developed to construct the null model of the test, and the prevalence degree of co‐location patterns is modeled as the significance level. The significance test is evaluated using the facility points‐of‐interest data sets. Experiments and comparisons show that the significance test can effectively detect network‐constrained spatial co‐location patterns with less priori knowledge and outperforms two state‐of‐the‐art methods in excluding spurious patterns.     

Reconstructing Population Density Surfaces from Areal Data: A Comparison of Tobler's Pycnophylactic Interpolation Method and Area-to-Point Kriging

We compare Tobler's pycnophylactic interpolation method with the geostatistical approach of area-to-point kriging for distributing population data collected by areal unit in 18 census tracts in Ann Arbor for 1970 to reconstruct a population density surface. In both methods, (1) the areal data are reproduced when the predicted population density is upscaled; (2) physical boundary conditions are accounted for, if they exist; and (3) inequality constraints, such as the requirement of non-negative point predictions, are satisfied. The results show that when a certain variogram model, that is, the de Wijsian model corresponding to the free-space Green's function of Laplace's equation, is used in the geostatistical approach under the same boundary condition and constraints with Tobler's approach, the predicted population density surfaces are almost identical (up to numerical errors and discretization discrepancies). The implications of these findings are twofold: (1) multiple attribute surfaces can be constructed from areal data using the geostatistical approach, depending on the particular point variogram model adopted—that variogram model need not be the one associated with Tobler's solution and (2) it is the analyst's responsibility to justify whether the smoothness criterion employed in Tobler's approach is relevant to the particular application at hand. A notable advantage of the geostatistical approach over Tobler's is that it allows reporting the uncertainty or reliability of the interpolated values, with critical implications for uncertainty propagation in spatial analysis operations.     

Assessing the Cluster Correspondence between Paired Point Locations

Some complex geographic events are associated with multiple point locations. Such events include, but are not limited to, those describing linkages between and among places. The term multi‐location event is used in the paper to refer to these geographical phenomena. Through formalization of the multi‐location event problem, this paper situates the analysis of multi‐location events within the broad context of point pattern analysis techniques. Two alternative approaches (Vector autocorrelation analysis and cluster correspondence analysis) to the spatial dependence of paired‐location events (i.e., two‐location events) are explored, with a discussion of their appropriateness to general multi‐location event problems. The research proposes a framework of cluster correspondence analysis for the detection of local non‐stationarities in the spatial process generating multi‐location events. A new algorithm for local analysis of cluster correspondence is proposed. It is implemented on a large‐scale dataset of vehicle theft and recovery location pairs in Buffalo, New York.     

Crime Risk Maps: A Multivariate Spatial Analysis of Crime Data

In crime analyses, maps showing the degree of risk help police departments to make decisions on operational matters, such as where to patrol or how to deploy police officers. This study statistically models spatial crime data for multiple crime types in order to produce joint crime risk maps. To effectively model and map the spatial crime data, we consider two important characteristics of crime occurrences: the spatial dependence between sites, and the dependence between multiple crime types. We reflect both characteristics in the model simultaneously using a generalized multivariate conditional autoregressive model. As a real‐data application, we examine the number of incidents of vehicle theft, larceny, and burglary in 83 census tracts of San Francisco in 2010. Then, we employ a Bayesian approach using a Markov chain Monte Carlo method to estimate the model parameters. Based on the results, we detect the crime hotspots, thus demonstrating the advantage of using a multivariate spatial analysis for crime data.     

Aggregation of Regional Economic Time Series with Different Spatial Correlation Structures. 具有不同空间相关结构的区域经济时间序列集成

In this article, we compare the relative efficiency of different forecasting methods of space‐time series when variables are spatially and temporally correlated. We consider two cases: (1) univariate forecasting (i.e., a space‐time series aggregated into a single time series) and (2) the more general instance of multivariate forecasting (i.e., a space‐time series aggregated into a coarser spatial partition). We extend the results in the literature by including the consideration of larger datasets and the treatment of edge effects and of negative spatial correlation. We first introduce a statistical framework based on the space‐time autoregressive class of random field models, which constitutes the basis of our simulation study, and we present the various alternative forecasting methods considered in the simulation. We then present the results of a Monte Carlo study related to univariate forecasting. In order to allow a comparison with the findings of Giacomini and Granger (2004), we consider the same forecasting strategies and the same combinations of the parameter values used there, but with a larger parametric set. Finally, we extend our analysis to the case of multivariate forecasting. The outcomes obtained provide operational suggestions about how to choose between alternative forecasting methods in empirical circumstances.     

Geostatistical Prediction and Simulation of Point Values from Areal Data

The spatial prediction and simulation of point values from areal data are addressed within the general geostatistical framework of change of support (the term support referring to the domain informed by each measurement or unknown value). It is shown that the geostatistical framework (i) can explicitly and consistently account for the support differences between the available areal data and the sought-after point predictions, (ii) yields coherent (mass-preserving or pycnophylactic) predictions, and (iii) provides a measure of reliability (standard error) associated with each prediction. In the case of stochastic simulation, alternative point-support simulated realizations of a spatial attribute reproduce (i) a point-support histogram (Gaussian in this work), (ii) a point-support semivariogram model (possibly including anisotropic nested structures), and (iii) when upscaled, the available areal data. Such point-support-simulated realizations can be used in a Monte Carlo framework to assess the uncertainty in spatially distributed model outputs operating at a fine spatial resolution because of uncertain input parameters inferred from coarser spatial resolution data. Alternatively, such simulated realizations can be used in a model-based hypothesis-testing context to approximate the sampling distribution of, say, the correlation coefficient between two spatial data sets, when one is available at a point support and the other at an areal support. A case study using synthetic data illustrates the application of the proposed methodology in a remote sensing context, whereby areal data are available on a regular pixel support. It is demonstrated that point-support (sub-pixel scale) predictions and simulated realizations can be readily obtained, and that such predictions and realizations are consistent with the available information at the coarser (pixel-level) spatial resolution.     

A Model for Mental Health Needs and Resourcing in Small Geographic Areas: A Multivariate Spatial Perspective

This paper considers how small area variations in a set of psychiatric referral outcomes in a London health authority of 750,000 people may inform health need assessment and health resourcing for mental illness based on true need. As well as adopting a multivariate perspective, the spatial interdependence of the outcomes is included in the modelling approach outlined. By contrast, existing studies on mental health need tend to focus on single outcomes, and may not include spatial dependence. The analysis relates to three hospital referral outcomes for psychiatric conditions, and to total community mental health referrals across sixty‐seven electoral wards in East London.     

Testing for Spatial Independence Using Similarity Relations

In this article, we construct new, simple, and nonparametric tests for spatial independence using symbolic analysis. An important aspect is that the tests are free of a priori assumptions about the functional form of dependence, making them especially suitable in situations where the dependence is nonlinear. We define the concept of a similarity relation, which is used to keep track of similarity between neighboring observations. This similarity count is used to construct new statistical tests based on both random permutation simulations and derived asymptotic distributions. We include a Monte Carlo study to better illustrate the properties and the behavior of the new tests under several synthetically generated processes. Apart from being competitive compared with other nonparametric and parametric tests, results underline the outstanding power of the new tests for nonlinear‐dependent spatial processes.     

Spatiotemporal Modeling of Correlated Small‐Area Outcomes: Analyzing the Shared and Type‐Specific Patterns of Crime and Disorder

This research applies a Bayesian multivariate modeling approach to analyze the spatiotemporal patterns of physical disorder, social disorder, property crime, and violent crime at the small‐area scale. Despite crime and disorder exhibiting similar spatiotemporal patterns, as hypothesized by broken windows and collective efficacy theories, past studies often analyze a single outcome and overlook the correlation structures between multiple crime and disorder types. Accounting for five covariates, the best‐fitting model partitions the residual risk of each crime and disorder type into one spatial shared component, one temporal shared component, and type‐specific spatial, temporal, and space–time components. The shared components capture the underlying spatial pattern and time trend common to all types of crime and disorder. Results show that population size, residential mobility, and the central business district are positively associated with all outcomes. The spatial shared component is found to explain the largest proportion of residual variability for all types of crime and disorder. Spatiotemporal hotspots of crime and disorder are examined to contextualize broken windows theory. Applications of multivariate spatiotemporal modeling with shared components to ecological crime theories and crime prevention policy are discussed.     

p‐Functional Clusters Location Problem for Detecting Spatial Clusters with Covering Approach

Regionalization or districting problems commonly require each individual spatial unit to participate exclusively in a single region or district. Although this assumption is appropriate for some regionalization problems, it is less realistic for delineating functional clusters, such as metropolitan areas and trade areas where a region does not necessarily have exclusive coverage with other regions. This paper develops a spatial optimization model for detecting functional spatial clusters, named the p‐functional clusters location problem (p‐FCLP), which has been developed based on the Covering Location Problem. By relaxing the complete and exhaustive assignment requirement, a functional cluster is delineated with the selective spatial units that have substantial spatial interaction. This model is demonstrated with applications for a functional regionalization problem using three journey‐to‐work flow datasets: (1) among the 46 counties in South Carolina, (2) the counties in the East North Central division of the US Census, and (3) all counties in the US. The computational efficiency of p‐FCLP is compared with other regionalization problems. The computational results show that detecting functional spatial clusters with contiguity constraints effectively solves problems with optimality in a mixed integer programming (MIP) approach, suggesting the ability to solve large instance applications of regionalization problems.     

Exploring scale-dependent correlations between cancer mortality rates using factorial kriging and population-weighted semivariograms

This article presents a geostatistical methodology that accounts for spatially varying population size in the processing of cancer mortality data. The approach proceeds in two steps: (1) spatial patterns are first described and modeled using population-weighted semivariogram estimators, (2) spatial components corresponding to nested structures identified on semivariograms are then estimated and mapped using a variant of factorial kriging. The main benefit over traditional spatial smoothers is that the pattern of spatial variability (i.e., direction-dependent variability, range of correlation, presence of nested scales of variability) is directly incorporated into the computation of weights assigned to surrounding observations. Moreover, besides filtering the noise in the data, the procedure allows the decomposition of the structured component into several spatial components (i.e., local versus regional variability) on the basis of semivariogram models. A simulation study demonstrates that maps of spatial components are closer to the underlying risk maps in terms of prediction errors and provide a better visualization of regional patterns than the original maps of mortality rates or the maps smoothed using weighted linear averages. The proposed approach also attenuates the underestimation of the magnitude of the correlation between various cancer rates resulting from noise attached to the data. This methodology has great potential to explore scale-dependent correlation between risks of developing cancers and to detect clusters at various spatial scales, which should lead to a more accurate representation of geographic variation in cancer risk, and ultimately to a better understanding of causative relationships.     

Effects of Confidentiality-Preserving Geo-Masking on the Estimation of Semivariogram and of the Kriging Variance

Geostatistical methods, such as semivariograms and kriging are well-known spatial tools commonly employed in many disciplines such as health, mining, forestry, meteorology to name only few. They are based essentially on point-referenced data on a continuous space and on the calculation of distances between them. In many practical instances, however, the exact point location, even if exactly known, is geo-masked to preserve confidentiality. This typically happens when dealing with confidential data related to individuals-health and their biometric parameters. In these situations, the estimation of the semivariogram and, hence, the spatial prediction can become biased and highly inefficient. This paper examines the extent of the bias in the particular case when the geo-masking mechanism is known (called “intentional locational error”) and lays the ground to a full understanding of the phenomenon in more general cases. We also examine how the geo-masking affects the estimation of the kriging variance thus reducing the efficiency of spatial prediction. We pursue our aims by developing some theoretical results and by making use of simulated and real data analysis.     

Spatial and social mobility

This paper analyzes the relationship between spatial mobility and social mobility. It develops a two‐skill‐type spatial equilibrium model of two regions with location preferences where each region consists of an urban area that is home to workplaces and residences and an exclusively residential suburban area. The paper demonstrates that relative regional social mobility is negatively correlated with segregation and inequality. In the model, segregation, income inequality, and social mobility are driven by differences between urban and residential areas in commuting cost differences between high‐skilled and low‐skilled workers, and also by the magnitude of taste heterogeneity.