Thank you to everyone reading. I would love a response ASAP! Licensing, tech, travel, and snow issues have massively delayed my project and my presentation is in 2.5 days from now 🙃
I am trying to do a regression to explore the relationship between human community statistics (demography, rurality, water use, etc) and the total amount of ecosystem service provision areas in Iowa, by county. In order to prepare my data for the regression, I did the following:
I joined a layer of Iowa county polygons to a table of candidate variables (note: for all joins, features were exported to a new layer to avoid issues). I merged a bunch of polygons into layers for different types of ecosystem services provision areas in Iowa to use as a dependent variable for the regression (Quality areas: merged polygons for protected lands, Supply areas: merged polygons for all public areas of natural landscape or greenspace, Impaired areas: merged layer of points of spill incidents, emission points, wastewater outfalls, etc). I then took these Iowa layers for provision areas and spatially joined them to their respective county (to the county polygon layer), creating a dataset with 99 records (for each IA county) containing the sum total of the provision area within the county, and all associated county demographic stats.
I used the shape area field for the polygons of county provision areas as the dependent variable for the exploratory regression. I chose my candidate variables, I believe I left all default options (not sure), and I ran the regression. Result: no significance, warning 001304 for the vast majority of possible models (WARNING 001304: Unable to estimate the following models due to severe multicollinearity (data redundancy).) Some models could be estimated, but the majority had the warning 001176: Perfect Multicollinearity. A relatively brief sample of the script at the end of the post. The full details/running script for the process is attached as a pdf.
I chose a variety of candidate variables, some of which could not possibly display perfect multicollinearity. I tried looking up help for this issue, but all that I could find recommended removing redundant variables until the issue is resolved (VIF decreases sufficiently), but the multicollinearity exists among almost all possible combinations of variables.
I feel like it might be because of how I did my joins, or maybe that I can't use shape area as a dependent variable, but I have very little idea what is wrong. I would love help on the exploratory regression, but I am also open to other data analysis ideas. If you have any questions I would be happy to answer.
Choose 3 of 12 Summary
Highest Adjusted R-Squared Results
AdjR2 AICc JB K(BP) VIF SA Model
0.31 3725.42 0.00 0.99 7.57 0.01 -HOUSEHOLDS_WITH_INDIVIDUALS_65_YEARS_AND_OVER*** +VACANT_HOUSING_UNITS*** +IRGCROPSCONSUMPTIVEUSE4CROPSFRESH_MGAL_D**
0.31 3726.49 0.00 0.96 7.93 0.01 -TOTAL_HOUSEHOLDS*** +VACANT_HOUSING_UNITS*** +IRGCROPSCONSUMPTIVEUSE4CROPSFRESH_MGAL_D**
0.30 3727.41 0.00 0.93 7.55 0.01 -TOTAL_POPULATION*** +VACANT_HOUSING_UNITS*** +IRGCROPSCONSUMPTIVEUSE4CROPSFRESH_MGAL_D**
Passing Models
AdjR2 AICc JB K(BP) VIF SA Model
WARNING 001304: Unable to estimate the following models due to severe multicollinearity (data redundancy).
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + MEDIAN_AGE__YEARS_ + TOTAL_HOUSEHOLDS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + WHITE + BLACK_OR_AFRICAN_AMERICAN.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + WHITE + TOTAL_HOUSEHOLDS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + WHITE + RENTER_OCCUPIED_HOUSING_UNITS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + BLACK_OR_AFRICAN_AMERICAN + TOTAL_HOUSEHOLDS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + HOUSEHOLDS_WITH_INDIVIDUALS_UNDER_18_YEARS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + HOUSEHOLDS_WITH_INDIVIDUALS_65_YEARS_AND_OVER.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + VACANT_HOUSING_UNITS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + RENTER_OCCUPIED_HOUSING_UNITS.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + RURALITY_INDEX.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + PSPERCAPITAUSE_GAL_PERSON_D.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + TOTAL_POPULATION + TOTAL_HOUSEHOLDS + IRGCROPSCONSUMPTIVEUSE4CROPSFRESH_MGAL_D.
WARNING 001176: Perfect Multicollinearity in: SHAPE_AREA ~ Intercept + WHITE + BLACK_OR_AFRICAN_AMERICAN + TOTAL_HOUSEHOLDS.
