Scenario: An epidemiologist (Katrina) is studying socio-economic status (SES) as part of a larger study of obesity, and assumes that she will detect positive spatial autocorrelation in neighborhoods of a large city by SES. Using her GIS she ranks her neighborhoods by whether median income is above or below the city median.

She examines the plot, which she finds convincing, but then decides to try some BW (Black-White) statistic to confirm her visual diagnosis, and get a measure of the strength of the spatial autocorrelation.

Questions:

  1. Is median income a better choice than mean income? Conjecture a distribution of SES, and consider whether it makes any difference.
  2. How should Katrina proceed to carry out the BW analysis? What might she need from her GIS?
  3. If Katrina has eight categories of SES, how might she best represent her data? What if she has SES for 10 different years?
  4. Suppose Katrina is missing data on one neighborhood (a brief rainfall washed away its value from her tablet as she wandered over the city measuring SES). How might she estimate the value of SES for this neighborhood? What are your tacit assumptions?
Return to Scenerios