I have calculated Moran's I in R and reject the null hypothesis in both data sets, there is evidence of strong spatial autocorrelation.

However, how do I compare the two?

First set:

Moran I statistic standard deviate = 9.7817, p-value < 2.2e-16

alternative hypothesis: greater

sample estimates:

Moran I statistic: 0.1736815445

Expectation: -0.0009970090

Variance: 0.0003188969

Second set:

Moran I statistic

standard deviate = 18.575, p-value < 2.2e-16

alternative hypothesis: greater

sample estimates:

Moran I statistic: 0.3307119647

Expectation: -0.0009970090

Variance: 0.0003189016

Both indicate a strong spatial autocorrelation, however, how can I precisely define between the two which one is stronger?

Is it simply by looking at which Moran's I value is greater?

Both have the same boundary (shapefile) and are in the same study area, but different time stamps.

1 Answer 1


Yes. If I_A > I_B for two data sets A and B, then there's greater spatial autocorrelation, where spatial autocorrelation is defined by the formula for the Moran I (other measures of spatial autocorrelation exist and may give different results). In short, neighbours of region i are more similar to i in data set A than in B, averaged all over.

Usually the next question asked is "But is it statistically significantly different?". That would require a statistical model for uncertainty in Moran's I, which I think can be computed under the assumption of no autocorrelation, but you've already rejected that for each data set. Moran's I is a statistic, not a random variable.

You could make it a random variable if you've got another hypothesis to test - for example suppose your measurements have a 5% uncertainty in them. That might not be so much that even at the extremes it affects the significance of the Moran's I test, but it might cause a difference in I_A and I_B to be down to the noise in the measurement. You could simulate 1000 noisy versions of A and B, compute 1000 I_A and I_B values and hence 1000 differences, then see where your actual difference ranks among those 1000 - this would be a Monte Carlo test of significance of the difference in spatial autocorrelation under the hypothesis of 5% noise in the measurements.

  • Excellent. Thanks for your thorough response.
    – perrier
    Dec 5, 2019 at 21:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.