How do we understand the efficiency levels between the Ordinary Least Square (OLS) Regression Model & Exploratory Regression (spatial statistics) when deriving models for spatial parameters related to urban form?
Generally speaking, OLS is the traditional statistic, which is a regression belonging to the ordinary linear regression model. It assumes that the relationship between input variables is constant over the whole study area, thsu the relationship can be regarded as a simple correlation analysis, such as Pearson correlation. Exploratory regression (spatial statistics) can be used to strengthen the input variables and results of OLS, see https://pro.arcgis.com/en/pro-app/2.7/tool-reference/spatial-statistics/exploratory-regression.htm.
If you would like to study the different relationship between variables (i.e. Urban Morphological), geographically weighted regression (GWR) is a better choice. It can study the varying relationships in different locations, with a simple implemention in ArcGIS, and has been widely used in urban sudies.
Generally speaking the dependant variable (DV) drives everything else in choice of regression method. Untill you mention your DV it's impossible to say what regression method you should use. The details are given in the attached screenshot. If you describe your DV we can help you in your choice. You can also follow the screenshot instructions. Please note that you need to repeat the Google search to get full information on this topic. Best wishes, David Booth
Hi
You can use OLS if the conditions are satisfied. These condition are linearity, No multicollinearity, hemogenity and no auto correlation between error.
Best regards
I thank Ahmad Al-shallawi , David Eugene Booth , and Haofan Xu for your valuable answers. I will keep your suggestions in mind while working on this. If anyone comes across any other insights related to this questions, do not hesitate to share.
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero