0
$\begingroup$

Is it possible to calculate a spatial regression with variables at different spatial levels WITHOUT any aggregation/disaggregation? Or is it necessary to consider the multilevel regression?

dependent variable (var): metric var at point level (x,y)
independent vars: nominal, ordinal, metric vars at different spatial levels (polygons), e.g. puffer air distance polygons from point; car, pedestrian isochrones from point; administrative levels (postcodes, counties, regions, states).

All of the vars refer to the points (x,y) which means that the points with the dependent var are located WITHIN the polygons or are centroids of these polygons (air buffer).

dataframe for regression calculation:

dependent var Y (x,y) | var A (postcode level) | var B (5 km buffer level) | ...  
Y1 | A1 | B1 | ...
Y2 | A2 | B2 | ...
...

aim: calculate local and global spatial regression models without considering multiple level spatial regression

I am using different global (Spatial Durbin, Spatial Error, ...) and local (geographically weighted regression - GWR) spatial models and have vars at different admin and non-admin (manually created) spatial levels. I know using a multiple level spatial regression would be better but since my spatial levels are not nested I would like to take variables from different levels in one regression and would like to know if this is statistically correct.

Thank you!

Improve this question
$\endgroup$
4
  • $\begingroup$ The details matter. Please explain how you propose to do such a regression. $\endgroup$
    – whuber ♦
    Commented Jul 21, 2023 at 21:21
  • $\begingroup$ @whuber question has been updated. $\endgroup$
    – Mapos
    Commented Jul 29, 2023 at 11:38
  • $\begingroup$ If you assume the spatial effects are additive, there's no need to set up a multi-level regression. $\endgroup$
    – jbowman
    Commented Jul 30, 2023 at 16:39
  • $\begingroup$ @jbowman what do you mean by additive spatial effects? All of the model vars as the standardized residuals show spatial autocorrelation (Moran I significant). Spatial effects of each var thus can influence the dependent var (spatial lags of mutliple vars are significant in global models). Can you please give an example of what you mean? Thank you! $\endgroup$
    – Mapos
    Commented Aug 1, 2023 at 5:31

0

Reset to default

Your Answer

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