2

I have a large point dataset of a population (By population, I mean that it was not a sample e.g. all houses in a county with an attribute such as size, price, age, etc.). Each point represents each case, their accurate position in space. It is possible to aggregate them up to any geographical unit such as census block, census tract, zip code etc. I would like to conduct several spatial statistical analyses (autocorrelation etc.) on this data. I tend to think that I should keep the data in the current format to maintain maximum variability rather than aggregating it to a spatial unit (e.g. census tract) to avoid MAUP and loss of variability.

However, is there ever a case to be made that under certain conditions a set of population point data should be aggregated to an areal unit?

5
  • To be clear: each point represents a settlement (town) and has a population value?
    – Martin F
    Mar 4, 2017 at 1:34
  • Also, do you have appropriate areal units for each point?
    – Martin F
    Mar 4, 2017 at 1:35
  • Each point represents each case, their accurate position in space. By population, I meant that it was not a sample. E.g. all houses in a county (with an attribute such as size, price, age, etc.). It is possible to aggregate them up to any geographical unit such as census block, census tract, zip code etc.
    – Naci
    Mar 4, 2017 at 2:15
  • I suggest you edit that new info into the question.
    – Martin F
    Mar 4, 2017 at 16:07
  • In general, computing statistics for data that is "aggregated up" will give you information about those aggregations (i.e. information about census blocks) rather than information about the individual points and their relationships. That said, this type of a question may be more appropriate for Cross Validated. Mar 5, 2017 at 23:19

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.