Prologue to Spatial Data Analysis in the Social Sciences

Introduction to spatial data analysis in the social sciences l.jpg
1 / 50
0
0
1293 days ago, 529 views
PowerPoint PPT Presentation
Session Objectives. Comprehend why spatial information investigation is importantIdentify sorts of inquiries for which SDA is relevantGain essential learning of the ideas, measurements, and systems for SDAIdentify some imperative issues and choice focuses inside SDALearn about a few assets for doing spatial information examination (programming, sites, books, and so forth.) Avoid losing all sense of direction in comparisons!.

Presentation Transcript

Slide 1

Prologue to Spatial Data Analysis in the Social Sciences RSOC597A: Special Topics in Methods/Statistics Kathy Brasier Penn State University June 14, 2005

Slide 2

Session Objectives Understand why spatial information examination is vital Identify sorts of inquiries for which SDA is applicable Gain fundamental learning of the ideas, insights, and strategies for SDA Identify some essential issues and choice focuses inside SDA Learn about a few assets for doing spatial information investigation (programming, sites, books, and so forth.) Avoid losing all sense of direction in conditions!

Slide 3

Why Do Spatial Analysis? "Everything is identified with everything else, except nearer things all the more so." (credited to Tobler)

Slide 4

Examples Is your instructive level liable to be like your neighbor's? Are homestead rehearses liable to be comparative on neighboring ranches? Are lodging values liable to be comparable in adjacent advancements? Do close-by neighborhoods have comparative theft rates?

Slide 5

County Homicide Rates 1990

Slide 6

What Is Spatial Data? 4 principle sorts occasion information, spatially persistent information, zonal information, spatial connection information Most often utilized as a part of sociologies is zonal information Data totaled to an arrangement of areal units (districts, MSAs, registration squares, ZIP codes, watersheds, and so forth.) Variables measured over the arrangement of units Examples: Census, REIS, County and City Databook, and so forth

Slide 7

What is Spatial Data Analysis? "The investigation of information on some procedure working in space, where techniques are tried to depict or clarify the conduct of this procedure and its conceivable relationship to other spatial marvels." Bailey and Gatrell (1995:7) Objective of spatial information examination: to comprehend the spatial course of action of variable qualities, identify designs, and analyze connections among factors

Slide 8

Why Do Spatial Data Analysis? To take in more about what you're contemplating To keep away from particular issues (missing factors, estimation blunder) To guarantee fulfillment of factual presumptions To be cool! To go insane! To take in more about measurements than you at any point needed or thought conceivable! To take in the confinements of measurements

Slide 9

Theoretical Reasons for Spatial Analysis It reveals to us something more about what we're considering Is there an unmeasured procedure that influences the marvel? Does this procedure show itself in space? Cases: connection forms, dispersion, recorded or ethnic legacy, automatic impacts

Slide 10

Statistical Reasons for Spatial Analysis Violation of relapse suspicions Units of investigation won't not be free Parameter appraisals are wasteful Estimated mistake difference is downwardly one-sided, which expands the watched R 2 values If spatial impacts are available, and you don't represent them, your model is not exact!

Slide 11

Examples of Research Using SDA Epidemiology (natural presentation explore) Criminology (wrongdoing designs) Education (neighborhood consequences for achievement) Diffusion/reception (innovations) Social developments (exchange unions, exhibitions) Market investigation (lodging and land value variety) Spillover impacts (monetary overflows of colleges) Regional reviews (territorial pay variety & disparity) Demography (isolation designs) Political science (decision thinks about)

Slide 12

BREAK!!

Slide 13

When do you have to do SDA? Is there a hypothetical motivation to speculate contrasts crosswise over space? Contrasts in wonders (variable qualities) Differences seeing someone between marvels (covariances) Are you utilizing information with spatial referent? On the off chance that yes to both, it is a smart thought to at any rate investigate any potential spatial impacts Exploration will reveal to you more about the subject you're contemplating

Slide 14

Spatial Independence Null speculation (H 0 ) Any occasion has an equivalent likelihood of happening at any position in the locale Position of any occasion is autonomous of the position of some other Implicit suspicion of much work in sociologies

Slide 15

Spatial Effects Test Hypothesis (H 1 ) Probability of an occasion happening not equivalent for every area inside district Position of any one occasion reliant on position of whatever other occasion Methods and measurements of SDA test this theory If upheld, can disclose to us more about what we're examining; can enhance our models If not bolstered, we realize that we have fulfilled suppositions

Slide 16

First Order Spatial Effects Non-uniform appropriation of perceptions over space Large-scale variety in mean over the spatial units Values of the factors are not free of their spatial area Results from communication of exceptional attributes of the units and their spatial area Ex: magnets and iron filings (Bailey & Gatrell) Referred to as spatial heterogeneity

Slide 17

Causes of Spatial Heterogeneity Patterns of social connection that make one of a kind qualities of spatial units Spatial administrations: legacies of provincial center outskirts connections => contrasts between units (pop, econ dvpt, and so forth.) Differences in physical components of spatial units Size of areas Combination: Differences in geology of units => distinctive examples of monetary improvement (extractive enterprises)

Slide 18

County Homicide Rates 1990 First request impacts?

Slide 19

Second Order Spatial Effects Localized covariation among means (or different insights) inside the district Tendency for intends to "take after" each other in space Results in bunches of comparative qualities Ex: magnets and iron filings (Bailey & Gatrell) Referred to as spatial reliance (spatial autocorrelation)

Slide 20

Causes of Spatial Dependence Underlying financial process has prompted to grouped conveyance of variable qualities Grouping forms gathering of comparative individuals in limited ranges Spatial association forms individuals close to each other more prone to interface, share Diffusion forms Neighbors gain from each other Dispersal forms People move, yet have a tendency to be short separations, bring their insight with them Spatial chains of importance Economic impacts that predicament individuals together Mis-match of process and spatial units Counties versus retail exchange zones Census piece bunches versus neighborhood systems

Slide 21

County Homicide Rates 1990 Second request impacts?

Slide 22

So now that I've persuaded you that spatial information investigation is an essential thought… . What Do We Do About It?

Slide 23

Goals of SDA To distinguish spatial impacts and their causes To suitably quantify spatial impacts To consolidate spatial impacts into models To enhance our insight into the procedure and how it happens over space All of these objectives require both hypothesis and strategies

Slide 24

Exploratory Spatial Data Analysis Start with inquiries regarding your hypothesis and information: Are there prone to be spatial procedures at work (dispersion, association, and so forth.)? Do your information units coordinate the procedure? (Messner et al. perusing) Visually and factually investigate your information Run essential distinct insights Map factors Look for examples, exceptions Look for spatial impacts (extensive scale variety, confined groups)

Slide 25

Gini Index 1989

Slide 26

How to Measure 'Space'? Need to characterize space keeping in mind the end goal to gauge its belongings Traditional ways (local sham factors, remove measures, and so forth.) Neighborhood structure Weights grid n x n framework, where: 0 = not neighbor 1 = neighbor

Slide 27

Weights Matrix "Neighbors" can be characterized as: Boundaries: Adjacent units (rook or ruler) Those units sharing some base/most extreme extent of normal limit Centroids If centroids are inside some predefined separate If unit would one say one is of k closest neighbors characterized by centroid separate Others? Choice to utilize one over another to some degree self-assertive Simpler is for the most part better Closer is by and large better Rely on hypothesis, your insight, and the ESDA to guide you

Slide 28

Weights Matrix Example Sample Region and Units Simple Contiguity (rook) Matrix

Slide 29

Statistical Tests for Spatial Dependence (Autocorrelation) Univariate Global Moran's I Indicates nearness and level of spatial autocorrelation among variable values crosswise over spatial units Where z is a vector of variable qualities communicated as deviations from the mean Where W is the weights lattice Expected estimation of I meetings on 0 when n is substantial; can do essentialness tests Large positive => solid grouping of comparative values Large negative => solid bunching of unique qualities

Slide 30

Global Moran's I and Moran Scatterplot Assesses relationship between the variable incentive for unit of source (x pivot) against the normal of the qualities its neighbors (y hub)

Slide 31

Local Indicators of Spatial Autocorrelation (LISA) Local Moran's I Decomposes worldwide measure into every unit's commitment Identifies the nearby 'hotspots', territories which contribute excessively to worldwide Moran's I

Slide 32

LISA Cluster Maps Homicide Rate 1990 Gini Index 1989

Slide 33

Additional Suggestions for ESDA Identify anomalies and hotspots both measurably and outwardly Try removing exception units from investigation and see what happens (does Moran's I change?) Explore changes in spatial examples after some time Compare (at least two) areas Split your specimen by a variable of premium Try diverse weights frameworks Play around with various covariates – get into your information!

Slide 34

BREAK!!!

Slide 35

Regression Modeling and SDA Use hypothesis and ESDA discoveries to create your model Procedure: Run OLS demonstrate Assess diagnostics If diagnostics show no spatial autocorrelation (or different infringement of relapse suppositions), OLS model is fine If diagnostics show spatial autocorrelation display, need to consider approaches to gauge and join spatial structure

Slide 36

OLS Diagnostics of OLS model will demonstrate kind of spatial impacts If either present, need to distinguish likely source Remedies Spatia

SPONSORS