Assessing Quality of Spatial Models Using the Structural Similarity Index and Posterior Predictive Checks

Model assessment is one of the most important aspects of statistical analysis. In geographical analysis, models represent spatial processes, where variability in mapped output results from uncertainty in parameter estimates. Slight spatial misalignments can cause inflated error scores when comparing maps of observed and predicted variables using traditional error metrics at the level of individual spatial units. We conceptualize spatial model assessment as a continuous value map comparison problem and employ methods from image analysis to score model outputs. The structural similarity index, a measure that attempts to replicate the human visual system using a local region approach, is used as an exploratory map comparison statistic. The measure is implemented within a Bayesian spatial modeling framework as a discrepancy measure in a posterior predictive check of model fit. Results are reported for simulation studies representing a variety of spatial processes in a spatial and space–time context. A case study of rainfall mapping in Sri Lanka demonstrates the proposed methodology applied to assessment of Bayesian kriging interpolations. Both simulation studies as well as the case study demonstrate that the approach reveals hidden spatial structure not uncovered by traditional methods. The spatially sensitive assessment methodology provides a diagnostic tool to support spatial modeling and analysis. La evaluación de modelos es uno de los aspectos más importantes de análisis estadístico. En el análisis geográfico, los modelos representan procesos espaciales en los que la variabilidad en los outputs es el resultado de la incertidumbre en los parámetros estimados. Leves desajustes espaciales pueden inflar los valores de error en la comparación entre los mapas de las observaciones y los mapas de las predicciones de las variables si es que se usan medidas tradicionales de medición de error al nivel de unidades espaciales individuales. Los autores conceptualizan la evaluación de modelos espaciales como un problema de comparación mapas de valor continuo y emplea métodos de análisis de imágenes para cuantificar los resultados del modelo. Se utiliza el índice de similitud estructural (SSIM), una medida que intenta replicar el sistema visual humano utilizando un enfoque de región local, como técnica de exploratoria comparación estadística de mapas. El índice es implementado dentro de un marco de modelización espacial bayesiano como medida de discrepancia en la comprobación posterior de predicción del desempeño del modelo. Los resultados se presentan para varios casos de simulación que representan una serie de procesos espaciales en un contexto espacio‐temporal y espacial. Un estudio de caso de mapeo de lluvias en Sri Lanka sirve como demostración de la metodología propuesta y su aplicación a la evaluación de las interpolaciones tipo krigeage (kriging) bayesianas. Tanto los estudios de simulación, así como el estudio de caso demuestran que el enfoque propuesto revela la estructura espacial oculta no evidenciada por métodos tradicionales. La metodología de evaluación espacialmente sensible que se presenta en este artículo proporciona una herramienta de diagnóstico para apoyar la elaboración de modelos y análisis espacial. 模型评估是统计分析中最为重要的内容之一。在地理分析中用模型表达空间过程,参数估计的不确定性会导致地图输出结果的可变性。当采用传统误差指标度量,在个体空间单位水平上进行观测和预测变量的地图比较时,微小的空间错位就可能导致误差的倍增。为此,本文通过将空间模型评估指标概念化为一个连续值图比较问题,并利用图像分析方法来评定模型输出。一种尝试以局域方法仿制人类视觉系统的度量指标——结构相似指数(SSIM),被用作为探索性地图的比较统计量。在贝叶斯空间模型框架下实现其量算,并将其作为一个偏差度量应用于模型拟合的后预测校验。仿真研究的结果显示出空间及时空环境下多类空间过程。以斯里兰卡降雨过程图为案例,展示了上述方法对贝叶斯克里格插值的准确性评估。仿真研究与实证结果均证明本文提出的方法可揭示以往传统方法掩盖的空间结构特征,空间敏感性评价为本研究的空间建模和分析提供了一个诊断工具。     

Beyond Moran's I: Testing for Spatial Dependence Based on the Spatial Autoregressive Model

The statistic known as Moran's I is widely used to test for the presence of spatial dependence in observations taken on a lattice. Under the null hypothesis that the data are independent and identically distributed normal random variates, the distribution of Moran's I is known, and hypothesis tests based on this statistic have been shown in the literature to have various optimality properties. Given its simplicity, Moran's I is also frequently used outside of the formal hypothesis-testing setting in exploratory analyses of spatially referenced data; however, its limitations are not very well understood. To illustrate these limitations, we show that, for data generated according to the spatial autoregressive (SAR) model, Moran's I is only a good estimator of the SAR model's spatial-dependence parameter when the parameter is close to 0. In this research, we develop an alternative closed-form measure of spatial autocorrelation, which we call APLE , because it is an approximate profile-likelihood estimator (APLE) of the SAR model's spatial-dependence parameter. We show that APLE can be used as a test statistic for, and an estimator of, the strength of spatial autocorrelation. We include both theoretical and simulation-based motivations (including comparison with the maximum-likelihood estimator), for using APLE as an estimator. In conjunction, we propose the APLE scatterplot, an exploratory graphical tool that is analogous to the Moran scatterplot, and we demonstrate that the APLE scatterplot is a better visual tool for assessing the strength of spatial autocorrelation in the data than the Moran scatterplot. In addition, Monte Carlo tests based on both APLE and Moran's I are introduced and compared. Finally, we include an analysis of the well-known Mercer and Hall wheat-yield data to illustrate the difference between APLE and Moran's I when they are used in exploratory spatial data analysis.     

Praying for Independence

The establishment of the Presbyterian Church of the New Hebrides in 1948 as an independent church was viewed by some participants as a step towards the independence of the nation, which occurred some 32 years later. This paper argues that the church was slow to promote an anticolonial perspective through the 1950s, though, as Indigenous clergy took on more senior roles in the church, there was a corresponding increase in political consciousness. The trans-colonial experiences of many young clergy – for education around the region or for meetings in the newly formed Pacific Conference of Churches in the 1960s – exposed participants to anticolonial theologies and the decolonising Pacific. When Indigenous clergy gained full control over the Presbyterian Church in 1973, they simultaneously demanded the end to the Condominium.     

A Typology of Spatial and Temporal Scale Relations

A simple typology of relations between any two geographical scales is established by qualitatively comparing their respective grains and extents. This typology is applied to spatial, temporal, and spatiotemporal scales. It describes seven relations between any two scales in either space or time. These basic relations yield a set of 169 qualitatively different spatiotemporal scale relations, a subset of which is portrayed dia‐grammatically. If it is possible to transform processes or patterns from one scale in the relation to the other, up to four scaling methods may need to be simultaneously applied, depending on the relation. Scaling methods might be classified as forms of grain generalization, grain decomposition, extent extrapolation, or extent selection. This typology may also provide a framework for investigations of dependencies between scales, as well as a reference scheme for observations of scale nonstationarity. The possibility is offered that any relation that forms a nonintersecting hierarchy in either space or time is a relation between essentially independent scales. However, the use of this typology is contingent on a number of factors, and it is offered as a tool, rather than a solution, for problems of scale.     

Large Sample-Size Distribution of Statistics Used In Testing for Spatial Correlation

Test statistics for testing for spatial correlation in continuous variables have been given by both Moran and Geary and have subsequently been generalized. It has been conjectured for a long time that under the hypothesis of no spatial correlations all these statistics are normally distributed when the sample size is large. This paper proves a very general theorem on the large sample normality of quadratic forms. As corollaries to the theorem the asymptotic normality, under the hypothesis, of all the above-mentioned statistics is established. The necessary conditions are quite unrestrictive. It is also shown, by means of a counter example, that the conditions given in a similar theorem (Cliff and Ord) are inadequate to ensure normality.     

Testing for Local Spatial Autocorrelation in the Presence of Global Autocorrelation

A fundamental concern of spatial analysts is to find patterns in spatial data that lead to the identification of spatial autocorrelation or association. Further, they seek to identify peculiarities in the data set that signify that something out of the ordinary has occurred in one or more regions. In this paper we provide a statistic that tests for local spatial autocorrelation in the presence of the global autocorrelation that is characteristic of heterogeneous spatial data. After identifying the structure of global autocorrelation, we introduce a new measure that may be used to test for local structure. This new statistic Oi is asymptotically normally distributed and allows for straightforward tests of hypotheses. We provide several numerical examples that illustrate the performance of this statistic and compare it with another measure that does not account for global structure.     

Exposing Unobserved Spatial Similarity: Evidence from German School Choice Data

In a spatial context, flexible substitution patterns play an important role when modeling individual choice behavior. Issues of correlation may arise if two or more alternatives of a selected choice set share characteristics that cannot be observed by a modeler. Multivariate extreme value (MEV) models provide the possibility to relax the property of constant substitution imposed by the multinomial logit (MNL) model through its independence of irrelevant alternatives (IIA) property. Existing approaches in school network planning often do not account for substitution patterns, nor do they take free school choice into consideration. In this article, we briefly operationalize a closed‐form discrete choice model (generalized nested logit [GNL] model) from utility maximization to account for spatial correlation. Moreover, we show that very simple and restrictive models are usually not adequate in a spatial choice context. In contrast, the GNL is still computationally convenient and obtains a very flexible structure of substitution patterns among choice alternatives. Roughly speaking, this flexibility is achieved by allocating alternatives that are located close to each other into nests. A given alternative may belong to several nests. Therefore, we specify a more general discrete choice model. Furthermore, the data and the model specification for the school choice problem are presented. The analysis of free school choice in the city of Dresden, Germany, confirms the influence of most of the exogenous variables reported in the literature. The estimation results generally indicate the applicability of MEV models in a spatial context and the importance of spatial correlation in school choice modeling. Therefore, we suggest the use of more flexible and complex models than standard logit models in particular. En un contexto espacial, los patrones sustitución flexible juegan un papel importante en el modelamiento del comportamiento de las decisiones individuales. Varios problemas de correlación pueden presentarse si dos o más alternativas de elección comparten características no observables por el modelador. Los modelos de valor extremo (multivariate extreme value‐MEV) ofrecen la posibilidad de relajar la propiedad de sustitución constante (constant substitution) presente en los modelos logit multinomiales (multinomial logit‐MNL), a través de su propiedad de independencia de alternativas irrelevantes (Independence of irrelevant alternatives property ‐IIA). A menudo, los enfoques existentes en la planificación de redes escolares no toman en consideración los patrones de sustitución y de libre elección de escuela. En este artículo, los autores presentan brevemente el funcionamiento de un modelo de elección discreta (discrete choice model) para la maximización de utilidad o modelo logit anidado generalizado (generalized nested logit model‐GNL) para dar cuenta de la autocorrelación espacial. Los autores sostienen que modelos demasiado simples y restrictivos no suelen ser adecuados en un contexto de elección espacial. En contraste el modelo GNL es conveniente en términos de su computación y obtiene una estructura muy flexible de los patrones de sustitución entre las alternativas de elección. En términos generales, esta flexibilidad se logra mediante la asignación (o anidación) de las alternativas cercanas en el espacio (una alternativa puede pertenecer a varios nidos). Por lo tanto, los autores presentan un modelo de elección discreta más general. El estudio presenta además datos y la especificación del modelo para un caso de elección de escuela concreto: el análisis de libre elección de escuela en la ciudad de Dresden, Alemania. El análisis confirma la influencia de la mayoría de las variables exógenas presentes en la literatura. Los resultados de la estimación demuestran en términos generales la aplicabilidad de los modelos MEV en un contexto espacial y la importancia de la autocorrelación espacial en el modelado de elección de escuela. Los autores concluyen sugiriendo el uso de modelos más flexibles y complejos que los modelos utilizados habitualmente, en particular los modelos logit estándar. 从空间视角看,灵活的替代模式在个人行为选择建模中发挥着重要作用。当存在两个或两个以上备选方案集具有共性且无法被建模者观察到时,就可能出现相关性问题。多元极值模型（MEV）通过不相关的替代属性（IIA）实现了对多元logit模型（MNL）中常数限制的松弛替代。现有校园网络规划方法通常无法解释替代模式,而且没有考虑到自由择校因素。本文简要地建立一个封闭离散选择模型（广义嵌套（GNL）模型）,从效用最大化角度来解释空间相关性。此外分析还表明,非常简单的约束模型通常不具有足够的空间选择情境。相比之下,GNL模型计算便捷,且可以在各选择方案中获得非常灵活的替代模式。大致而言,这种灵活性大体是通过与住处位置距离上彼此靠近的替代选择分配而获得,一个给定的选择可能属于不同的住处。因此,我们给出了一个更一般的离散选择模型。此外,还给出了针对择校问题的数据和模型设定。基于德国德累斯顿市自由择校分析,证实了已有研究中多数外生变量的影响。估计结果证实了MEV模型在空间分析中的适用性以及择校模型中空间相关的重要性,并建议使用更加灵活和复杂的模型而不是标准的logit模型。     

Exploring Spatial Patterns Using an Expanded Spatial Autocorrelation Framework

Spatial autocorrelation (SA) is regarded as an important dimension of spatial pattern. SA measures usually consist of two components: measuring the similarity of attribute values and defining the spatial relationships among observations. The latter component is often represented by a spatial weights matrix that predefines spatial relationship between observations in most measures. Therefore, SA measures, in essence, are measures of attribute similarity, conditioned by spatial relationship. Another dimension of spatial pattern can be explored by controlling observations to be compared based upon the degree of attribute similarity. The resulting measures are spatial proximity measures of observations, meeting predefined attribute similarity criteria. Proposed measures reflect degrees of clustering or dispersion for observations meeting certain levels of attribute similarity. An existing spatial autocorrelation framework is expanded to a general framework to evaluate spatial patterns and can accommodate the proposed approach measuring proximity. Analogous to the concept of variogram, clustergram is proposed to show the levels of spatial clustering over a range of attribute similarity, or attribute lags. Specific measures based on the proposed approach are formulated and applied to a hypothetical landscape and an empirical example, showing that these new measures capture spatial pattern information not reflected by traditional spatial autocorrelation measures.     

Nonparametric Estimation of the Spatial Connectivity Matrix Using Spatial Panel Data

We use moments from the covariance matrix for spatial panel data to estimate the parameters of the spatial autoregression model, including the spatial connectivity matrix W. In the unrestricted spatial autoregression model, the parameters are underidentified by one when W is symmetric. We show that a special case exists in which W is asymmetric and its parameters are exactly identified. If the panel data are stationary and ergodic, spatially and temporally, the estimates of W and the spatial autoregression coefficients are consistent. Spatial panel data for house prices in Israel are used to illustrate this methodology. Los autores usan momentos de una matriz de covarianza para datos panel espaciales para estimar los parámetros del modelo de autoregresión espacial (spatial autoregressive model), incluyendo la matriz de conectividad (o de ponderación) espacial W. En el modelo de autoregresión espacial sin restricciones, los parámetros están sub‐identificados por un valor de uno en los casos que la matriz W es simétrica. Los autores demuestran que existe un caso especial en el cual la matriz W es asimétrica y sus parámetros tienen cálculo exacto. Si los datos panel son estacionarios y ergódicos, espacial y temporalmente, los estimados de W y el coeficiente de autoregresión espacial son consistentes. Para ilustrar la metodología propuesta, los autores usan datos‐panel espaciales de precios de vivienda en Israel. 本文通过采用空间面板数据的协方差矩阵对包含空间相关矩阵W的空间自回归模型进行参数的矩估计。在无约束空间自回归模型中,W是对称矩阵时,参数可由其估计得到。本文展示了一种W是对称矩阵且其参数能够被精确估计的特殊情况。如果面板数据在时间与空间特征上具有平稳性和遍历性,那么W和空间自回归参数的估计是一致的。最后,针对以色列住房价格的空间面板数据采用此方法进行实证研究。     

Using AMOEBA to Create a Spatial Weights Matrix and Identify Spatial Clusters

The creation of a spatial weights matrix by a procedure called AMOEBA, A Multidirectional Optimum Ecotope-Based Algorithm , is dependent on the use of a local spatial autocorrelation statistic. The result is (1) a vector that identifies those spatial units that are related and unrelated to contiguous spatial units and (2) a matrix of weights whose values are a function of the relationship of the ith spatial unit with all other nearby spatial units for which there is a spatial association. In addition, the AMOEBA procedure aids in the demarcation of clusters, called ecotopes, of related spatial units. Experimentation reveals that AMOEBA is an effective tool for the identification of clusters. A comparison with a scan statistic procedure (SaTScan) gives evidence of the value of AMOEBA. Total fertility rates in enumeration districts in Amman, Jordan, are used to show a real-world example of the use of AMOEBA for the construction of a spatial weights matrix and for the identification of clusters. Again, comparisons reveal the effectiveness of the AMOEBA procedure.     

"台独"势力的"日本情结"问题

“台独”分子有一个共同的特性,就是都具有浓厚的“日本情结”,这也是驱动“台独”势力长期不遗余力从事分裂祖国活动和造成台湾问题久拖不决的重要因素。究其原因,主要有二：1、日本对台湾50年的殖民统治,特别是它所发动的“皇民化运动”的久远影响,是导致“台独”势力“日本情结”产生的历史原因;2、“台独”势力出于台湾“独立”的欲求,急需取得外部势力特别是与台湾渊源甚深的日本右翼势力的支持,这是“台独”势力“日本情结”产生的现实原因,也是主要原因。     

Exploring Goodness-of-Fit and Spatial Correlation Using Components of Tango's Index of Spatial Clustering

The ability to detect anomalies such as spatial clustering in data sets plays an important role in spatial data analysis, leading to interest in test statistics identifying patterns exhibiting significant levels of clustering. Toward this end, Tango (1995) proposed a statistic (and its associated distribution under a null hypothesis of no clustering) assessing overall patterns of spatial clustering in a set of observed regional counts. Rogerson (1999) observed that Tango's index may be decomposed into the summation of two distinct statistics, the first mirroring standard tests of goodness-of-fit (GOF), and the second an index of spatial association (SA) similar to Moran's I . In this article, we investigate the effectiveness of Rogerson's expression of Tango's statistic in separating GOF from SA in data sets containing clusters. We simulate data under the null hypothesis of no clustering as well as two alternative hypotheses. The first alternative hypothesis induces a poor fit from the null hypothesis while maintaining independent observations and the second alternative hypothesis induces spatial dependence while maintaining fit. Using Rogerson's decomposition and leukemia incidence data from upstate New York, we show graphically that one is unable to statistically distinguish poor fit from autocorrelation.     

Testing for Regional Differences in Means: Distinguishing Inherent from Spurious Spatial Autocorrelation by Restricted Randomization

Tests for differences among regional means are typically carried out by analysis of variance (ANOVA). When such data are spatially autocorrelated (SA), the assumptions of ANOVA are not met, giving rise to excessive type I error rates. Two spatially adjusted ANOVA methods, Griffith's and COCOPAN, have been proposed to overcome this problem. In this study we show, by means of extensive simulations, the magnitude of the error rates introduced by SA induced in isolation-by-distance models typical of those used in population genetics. For data suspected of exhibiting such SA, we propose a strategy for distinguishing between inherent SA, generated within the data by a contagious process, and spurious SA, introduced by regional differences in means. The approach adopted is that of restricted randomization of distance matrices. We also furnish error rates and power estimates for both Griffith's method and COCOPAN. In addition to the simulated data, the methods are applied to an actual example from plant population biology.     

Using Spatial Filters and Exploratory Data Analysis to Enhance Regression Models of Spatial Data

Residual spatial autocorrelation is a situation frequently encountered in regression analysis of spatial data. The statistical problems arising due to this phenomenon are well‐understood. Original developments in the field of statistical analysis of spatial data were meant to detect spatial pattern, in order to assess whether corrective measures were required. An early development was the use of residual autocorrelation as an exploratory tool to improve regression analysis of spatial data. In this note, we propose the use of spatial filtering and exploratory data analysis as a way to identify omitted but potentially relevant independent variables. We use an example of blood donation patterns in Toronto, Canada, to demonstrate the proposed approach. In particular, we show how an initial filter used to rectify autocorrelation problems can be progressively replaced by substantive variables. In the present case, the variables so retrieved reveal the impact of urban form, travel habits, and demographic and socio‐economic attributes on donation rates. The approach is particularly appealing for model formulations that do not easily accommodate positive spatial autocorrelation, but should be of interest as well for the case of continuous variables in linear regression.