Methods of analysing change over time and space

1
Methods of analysing change over time and space

• Ian Gregory (University of Portsmouth)
• Paul Ell (Queens University, Belfast)

2
Advantages of temporal GIS data

• 1. Integration
• Potentially any data with a spatial and a
  temporal reference can be integrated
• Allows new data to be created
• 2. Analysis
• Need to spot broad trends and places/times that
  show different patterns
• Only limited techniques available
• Multi-level modelling
• Geographically Weighted Regression (GWR)
• 3. Visualisation
• Allows exploration of data and presentation of
  results
• In all cases we want to make best use of all of
  the available detail in the data (attribute,
  spatial and temporal)

3
Data integration District-level net migration
rates

• Net migration from the basic demographic
  equation
• NMt,tn (ptn pt) - (Bt,tn - Dt,t
  n)
• Age and sex specific population, fertility and
  mortality data have been published decennially in
  Britain since the 1850s
• Net migration for women aged 5 to 14 at the start
  of the decade can be calculated as
• Females aged 15 to 24 at end of decade minus
  females aged 5 to 14 at start of decade minus
  number of deaths in the cohort through the decade
• Problem As net migration is the residual it is
  highly susceptible to error. In particular, the
  impact of any boundary changes will appear as
  migration.
• Traditional studies
• Most studies of net migration use county-level
  data to avoid boundary change issues
• Only use the census so are unable to sub-divide
  migrants by age/sex

4
Net migration through areal interpolation

• Standardise population and mortality data from
  many dates onto a single set of target units
• Integrate data from census and Registrar
  Generals Decennial Supplement
• Allows us to calculate net migration rates for
  males and females in ten-year cohorts from ages 5
  to 14 to ages 55 to 64 (at start of decade).

5

Standardised time-series
Net migration rates among the 5 to 14 cohort
Bristol
Cheltenham
Westbury
6
Detailed attribute comparisons
Net migration rates among different cohorts in
the 1920s
Bristol
Cheltenham
Westbury
7
Net migration strengths and weaknesses

• Strengths
• From the census (comprehensive)
• Can compute complete time-series from 1851-2001
• Can be integrated with other aggregate
  information
• Pop. density
• Employment
• Social class
• Proximity to coast/areas of natural beauty, etc.
• Weaknesses
• No information on flows
• Low net mig. can be caused by high in and out
  mig. cancelling each other out
• Ecological fallacy when analysing data

8
Other sources

• Pooley Turnbull (1996)
• Sample of 75,000 migrations by 16,000 people born
  1750-1930 created using genealogical societies.
• Gives
• Where each move was to and from (including grid
  references)
• When the move occurred
• Large amounts of attribute information on
  employment, family structure, etc.
• Strengths
• Detailed individual-level info
• Weaknesses
• Potentially biased sample
• Doesnt include the young up to the present

9
Bringing them together

• Both datasets are geo-referenced can be
  integrated
• Allows
• Comparison of individual-level and ecological
  data (use of multi-level modelling)
• Tests whether ecological and individual level
  relationships are consistent
• Evaluates the accuracy of the sample
• Therefore
• Integrates different datasets
• Makes full use of spatial, attribute and temporal
  information

10
Spatial analysis with GWR

• Global vs local analysis
• Global analysis
• Gives a single summary statistic or equation for
  whole study area
• Average relationship implies spatial
  homogeneity
• Local analysis
• Allows parameters to vary over space
• Shows how relationships vary geographically
• Allows spatial heterogeneity

11
Geographically Weighted Regression

• Descriptive Allows the relationship between the
  variables to vary over space by providing
  separate intercept and regression coefficients
  for every location on the map
• Test as to whether the model shows significant
  spatial variation
• Conventional regression yia0a1x1ia2x2iei
• GWR yia0(ui,vi)a1(ui,vi)x1ia2(ui,vi)x2iei
• (ui,vi) represents the coordinates of the ith
  point and an(ui,vi) is the impact of an(u,v) at
  the ith point. This is implemented using a
  distance decay model

12
Example

• Global LTLIi3.896.6UNEMi31.1CROWi-3.5SPFi-22.5
  SC1i-5.6DENSi

Intercept
SC1
UNEM
DENS
13
Mapping the R2i values
Source Fotheringham et al, 1998
14
Uses in spatio-temporal analysis

• In C19 young women migrated as much as men but
  the spatial pattern differed significantly
  because of the different employment opportunities
  (main employers domestic service, textiles)
• Conventional regression
• Mig proportional to DS and Text
• GWR
• Textiles attract women in Lancs/W. York
• DS attracts women to wealthy areas eg West
  London, Cheltenham, Leamington Spa
• Over time this pattern will become more complex
  and the differences between men and women will
  reduce

15
(No Transcript)