Intro Flashcards
What is Spatial Analysis?
- A set of tools (stats, math, software, hardware) to analyze (concepts, theories, techniques, models) spatial processes
- A subset of analytic techniques whose result depend on the spatial frame, or will change if the frame changes or if objects are repositioned within it.
- Reveals patterns that are otherwise invisible
Aspatial data
Attribute, Pi (z)
- Value/info
Spatial Data
Location and Attribute
- Pi (x,y,z)
- This is why matrices are often used
What does it mean that Spatial Analysis has no locational invariance?
- Results change when locations of study objects change
- ‘Where’ matters
What are the 4 main components of spatial analysis?
- Data manipulation
- Exploratory spatial data analysis and visualization
- Spatial statistical analysis
- Spatial modelling
Data manipulation
- GIS, databases, processing, projecting
Exploratory spatial data analysis and visualization
Showing and identifying interesting patterns
Spatial statistical analysis
Investigating data to determine whether or not data can be represented in spatial model
Spatial Modeling
Explaining interesting patterns and/or predict spatial outcomes
Spatial Sampling
- Location as an experimental design problem
- Location as a given
Location as an experimental design problem
- Spatial sampling = where to collect data
- Which villages
- Where to locate air quality monitoring stations
- Design sampling approach to fit surface
Location as a given
- Most spatial data analyses have no choice in location
- No sampling in the usual sense
- data = attributes augmented with location information
- ex. census tract boundaries not under control of analyst
Spatial Autocorrelation
- Why is something the way it is?
- There is an underlying process for why the surface is the way it is (not random as is ‘required’ for stats)
- Ex. Elevation has underlying trend in topography, tectonics, erosion
What are the 4 major problems in spatial sampling?
- Maup
- Ecological fallacy
- Boundary/extent
- Scale
What is critical when ‘the where’ is introduced?
- Spatial dependence, the relatedness of data in space
MAUP
- Modifiable Areal Unit Problem
- Problem when data relates to discrete zones (most socioeconomic data)
- Densities in an area change/vary in space
- Grouping the data can have infinite possibilities and can greatly affect results (mean, etc.)
- Data is strongly dependent on groupings (tell different stories)
- Partly depends on underlying micro data and nature of zoning system
- Some can be justified (watersheds) but some change over time (neighbourhoods)
- Paly depends on underlying micro data and nature of zoning system
Correlation relationship, R^2
- Does not imply causation
- Shows how strong the relationship is between the dependent and independent variables
- Can change/shift based on aggregation of groups
Scale Problem
- Scale of spatial process and scale of spatial measurement
- Up/down scale and results change
- Use fractals (similar spatial pattern at increasing scales) to understand how to scale up/down
Ecological Fallacy
- Inference on individual based on aggregated group data
- Results from belief that relationships observed fro groups hold for individuals
- Ex. use aggregated province to infer on municipalities, can greatly differ from provincial mean
- Ex. Countries with more fat in diet have higher rate of breast cancer, must mean women who eat fatty foods more likely to get cancer
Boundary Problem
- Spatial processes are generally unbounded
- Artificial and arbitrary boundaries are often imposed for analysis purposes (grid points)
- Edge effects outside of boundary often impossible to control
- Does surface extend outside study area even though we have no observations?
- Potential fix: collect data outside of study area to help control edge effects
Diffusion
- Who has it, who doesn’t
- Spreads slowly outwards
- Requires contact/adjacency
Exchange and Transfer
- Commodities and income
- Adjacency
- Spill over effects
Interaction
- Events at a location affect events at another
Spatial organization can be exploited to?
- Design sampling plans
- Interpolate
- Fill in missing values in a database
- Classify
Main issues of Spatial Processes?
Representation of spatial dependence
Spatial Processes
- Diffusion
- Exchange and Transfer
- Interaction
Spatial Dependence
- Recall 1st law of geography
- Affects outcome of stat tests
- If present and not accounted for, the variance of correlation coefficient is underestimated
- Overlap on graphs
- Leads to redundancy, greater chance of outliers, chance of accepting null hypothesis when it is wrong
1st Law of Geography
Everything is related to everything else, but near things are more related than distant things
How do you tell if data is spatially dependent?
- Test for it!
- Geary’s C and Moran’s I (Positive, Negative, None)
First order spatial autocorrelation
- Spatial variation occurs when observations across a study region vary from place to place due to changes in the underlying properties of the local environment
Second order spatial autocorrelation
Variation is due to interaction effects between observations