communicating meaning and purpose of spatio-temporal data analysis (master's thesis)
TRANSCRIPT
Communicating meaning and purposeof spatio-temporal data analysis
Master thesis by Matthias Hinz
Institute for Geoinformatics (ifgi)University of Mnster
July 15h, 2016
Spatio temporal analysis tries to explain the real world based on data and hypothetical processes that generate data
Explanation, prediction, problem solving
Purpose (Why?, Who?)
How?
Meaning
What? Where? When?
Therories, models, information
Data & hypothetical processes
Interpretation, analysis
Simple Definition of purpose
: the reason why something is done or used : the aim or intention of something
: the feeling of being determined to do or achieve something
: the aim or goal of a person : what a person is trying to do, become, etc.http://www.merriam-webster.com/dictionary/purpose
Simple Definition of meaning
: the idea that is represented by a word, phrase, etc.
: the idea that a person wants to express by using words, signs, etc.
: the idea that is expressed in a work of writing, art, etc.
http://www.merriam-webster.com/dictionary/meaning
The way in which we identify instances of water hangs on the capacity to identify its hidden structure, its essential properties, and not just its superficial characteristics such as color and odorBortolotti, L, An introduction to the Philosophy of Science, p. 85
- Spatio-temporal Point data set- Region of Northcumbria, EnglandMight we interpolate?(e.g. Space-time-kriging?)
- Round, red, organic- Maybe eatable
Might be a tomato?
- Spatial Point data set (fixed time)- Meuse region, NetherlandsMight we interpolate?
stpp: An R Package for Plotting, Simulating andAnalyzing Spatio-Temporal Point Patterns
Might we interpolate?No, because the data refers to outrbreaks of the foot-and-mouth disease, which are discrete Events(Spatio-temporal point pattern)
Might we interpolate?Yes, because the data observed from a continuous Spatial Field. (Zinc concentration in the river floodplain)Unknown data can be estimated
Technical descriptions (and visualization) are not sufficient for meaningful interpretation and further analysis
stpp: An R Package for Plotting, Simulating andAnalyzing Spatio-Temporal Point Patterns
Modelling spatio-temporal information generation By Simon Scheider, Benedikt Grler, Edzer Pebesma, and Christoph Stasch (2016) Propose an algebraic model that describes possible spatiotemporal data derivations as functions on data generation procedures
Introduce spatio-temporal information types in terms of reference systems
Describe how information is generated, including raw observations as well as derived products
Modelling spatio-temporal information generation By Simon Scheider, Benedikt Grler, Edzer Pebesma, and Christoph Stasch (2016)
Image source: Scheider et al
Communicating meaning and purpose
of spatio-temporal data analysis
(Master thesis)
Provides a prove-of-concept of the proposed algebraic model:Shows how to use R (resp. Rho) with selected use-cases of spatio-temporal analysis and generate spatio-temporal derivation graph enriched with algebraic semantics
Facilitates meaningful interpretation, comprehension and validation of an spatio-temporal analyses
Suggests how to communicate meaning and purpose
Communicating meaning and purpose
of spatio-temporal data analysis
(Master thesis)
Automated workfow:1. Execute script and perform analysis in Rho (formerly CXXR, supports provenance tracking)2. Retrieve provenance information and create derivation graph3. Enrich graph with algebraic semantics4. Visualize/export graph with Rgraphviz(5. Further work: publish and and query graph with semantic web tools SPARQL, RDF, OWL...)
Use Case: Interpolation of zinc concentrations in the ground
1. Execute Analysis
demo(meuse, ask=FALSE, echo=FALSE)meuse$lzinc = log(meuse$zinc)zincPointData = SField(meuse["lzinc"], meuse.area)interpolator = getInterpolator(modelSemivariogram(zincPointData), zincPointData)locInterest = SField(geometry(meuse.grid), geometry(meuse.grid), cellsArePoints = TRUE)intZincPointData = interpolator(locInterest)spplot(intZincPointData@observations[1])
R script (shortened):
Interpolated zinc values (output):
2-4. Derivation graph
Data / literal values (incl. unevaluated functions)
operation (incl. function calls)
Data is fraction of other data / Data is input of an operation
Operation produces output
Function is called by an operation
2. Derivation graph: Data acquisition
2. Derivation graph: locations of interest
Note that function name SField is misleading and will be revised later on.
2. Derivation graph: interpolation
Evaluation and Conclusion
It is possible to generate spatio-temporal derivation graphs that enhance understanding and comprehension of an analysis, thus enable better communication
Applicability and scale of the algebra has yet to be proven
The final results will be discussed and matched with requirement of comprehensible science
Thank you for your attention!
Any questions?
"He who loves practice without theory is like the sailor who boards ship without a rudder and compass and never knows where he may cast"Leonardo Da Vinci
Resources
Meaningful Spatial Statistics: http://meaningfulspatialstatistics.org/
The Rho project (formerly CXXR, includes Provenance tracking for R): https://github.com/rho-devel/rho
Graphviz - Graph Visualization Software: http://www.graphviz.org/
Rgraphviz, Interfaces R with the AT and T graphviz library for plotting R graph objects from the graph package: http://bioconductor.org/packages/release/bioc/html/Rgraphviz.html
Sources
Scheider, Simon, et al. "Modeling spatiotemporal information generation." International Journal of Geographical Information Science, 2016, p. 1-29.
Christopher Anthony Silles. Provenance-Aware CXXR. PhD thesis, University of Kent, 2014.
Gabriel, Edith; Rowlingson, Barry; Diggle, P. stpp: an R package for plotting, simulating and analyzing Spatio-Temporal Point Patterns. Journal of Statistical Software, 2013, 53. Jg., Nr. 2, S. 1-29.
Bortolotti, L. "An introduction to the philosophy of science", 2008.
Appendix (another graph still under developement)