When to address spatial auto-correlation?

by Steve_s   Last Updated September 17, 2018 18:19 PM

I am trying to understand when I should address spatial auto-correlation.

Let's say that I have a number of weather stations in the mountains and an equal number of weather stations by the sea. I want to measure if there is a statistically significant difference in annual rainfall between these two groups.

I am not trying to build any predictive model. I just want to perform a t-test to compare the two situations.

The observations within the two groups are not independent:

  • Each group comes from a relatively uniform area (mountains vs. seaside).
  • Each area has distinct but relatively uniform weather patterns.
  • Changes in weather within an area would be detected with no lag by all stations in the same area.

Nonetheless, all stations in a groups are not identical copies. Each one of them measures unique rainfall values due to very local topography, variability in wind patterns, etc.

Is spatial auto-correlation an issue in this situation?

Related Questions

Significance of Spatial Data

Updated November 24, 2017 02:19 AM

determining the significance of change

Updated May 08, 2017 19:19 PM