Classifying Health-related Data for Disease...
Transcript of Classifying Health-related Data for Disease...
Classifying Health-related Data for Disease Mapping
IRWIN F. CHAVEZ
FACULTY OF TROPICAL MEDICINE
MAHIDOL UNIVERSITY
APRIL 24, 2018 THAI GI S NET MEETUP SERIES
Map of zoonotic pathogens from wildlife, shown from lowest occurrence (green) to highest (red)SOURCES: http://www.columbia.edu/cu/news/08/02/hotspots.html; https://www.nature.com/articles/nature06536.pdf)
In disease mapping…Data can be presented in an infinite number of ways
Mapping paradox◦ to show something, you have to “hide” something
◦ How to Lie with Maps by Mark Monmonier◦ Mapmakers are compelled to tell “white lies”
◦ Multitude of ways to “lie”
◦ The “lies” are either a) necessary, b) deliberate, c) unintended
District-level map of Thailand
926 features
Equal intervalIntervals are based on the maximum value divided by the number of intended classes
EXAMPLE:
Value range: 0 to 600
Classified into 5 groups◦ 0 to 120
◦ 121 to 240
◦ 241 to 360
◦ 361 to 480
◦ 481 to 600
DOES NOT take into account data distribution
Equal interval map of DF/DHF cases876 features; 50 missing data
652/5 = 131 class width
Number of features per category
Natural breaksAlso called Jenks’ optimization method
Organizes data into “natural groups”
Iterative computations until lowest in-class deviations are achieved
Takes into account data distribution and variability and uses those to identify “natural” categories
Map of DF/DHF cases using Jenks natural breaks876 features; 50 missing data
Standard deviationBased on the data’s mean and standard deviation
Takes into account data distribution and dispersion in building class boundaries
Standard deviational map of DF/DHF cases876 features; 50 missing data
Mean = 44.5; SD = 66.3
mean
QuantileBased on ranks
4 – quartile; 5 – quintile; 10 – decile; 100 – centile
Reduces the effect of outliers
Quantile classification of DF/DHF cases876 features; 50 missing data
5 categories (quintiles)
QuantileBased on ranks
4 – quartile; 5 – quintile; 10 – decile; 100 – centile
WAIT… a few considerations◦ Nature of data
◦ Is there ranking?
◦ Tied observations
Quantile classification of DF/DHF incidence rate per 10,000 population876 features; 50 missing data
5 categories (quintiles)
QuantileBased on ranks
4 – quartile; 5 – quintile; 10 – decile; 100 – centile
WAIT… a few considerations◦ Nature of data
◦ Is there ranking?
◦ Tied observations
Not suitable for counts
Custom classificationCounts of reported malaria cases in Mekong countries at administrative level 1 features
SOURCE: Mekong Malaria II Update of malaria, multi-drug resistance and economic development in the Mekong region of Southeast Asia; The Southeast Asian Journal of Tropical Medicine and Public Health, vol 34 suppl 4, 2003.
Custom classification“Malaria control era”
Species ratio highlights predominating species at administrative level 1 features in the Mekong region (2000-2001)
SOURCE: Mekong Malaria II Update of malaria, multi-drug resistance and economic development in the Mekong region of Southeast Asia; The Southeast Asian Journal of Tropical Medicine and Public Health, vol 34 suppl 4, 2003.
Map of zoonotic pathogens from wildlife, shown from lowest occurrence (green) to highest (red)SOURCES: http://www.columbia.edu/cu/news/08/02/hotspots.html; https://www.nature.com/articles/nature06536.pdf)
Take home messagesKnow and understand options to present data
Be critical
GIS is data-centric – keep the statistician in you awake
Be creative in designing maps that optimize data use, but not at the expense of clarity
Don’t worry about the “lies” in your maps – stay true to your map’s objective(s)
… THANK YOU!
LINKS/REFERENCESMonmonier, Mark. (2005). Lying with Maps. Statistical Science. 20. 10.1214/088342305000000241. (https://goo.gl/RX1U6u)
Mitchell, Andy. (1999). The ESRI® Guide to GIS Analysis Vol 1: Geographic Patterns & Relationships. Environmental Systems Research Institute, Inc.
Krygier, John and Wood, Denis. (2005). Making Maps: A Visual Guide to Map Design for GIS. The Guildford Press.
Singhasivanon, Pratap et al. (2003). Mekong Malaria II Update of malaria, multi-drug resistance and economic development in the Mekong region of Southeast Asia. The Southeast Asian Journal of Tropical Medicine and Public Health Volume 34 Supplement 4, 2003.
http://www.columbia.edu/cu/news/08/02/hotspots.html
https://www.nature.com/articles/nature06536.pdf (MU access required to download)
http://open.lib.umn.edu/mapping/chapter/7-lying-with-maps/
https://gisgeography.com/choropleth-maps-data-classification/
https://en.wikipedia.org/wiki/Jenks_natural_breaks_optimization