Lets say there are 8 explanatory variables which and 1 dependent variable. These are both tested with an OLS and GWR model and AICc values are obtained. Say 2 and 1 respectively. I understand that the lower AICc value has a better fit and model performance and thus if the GWR AICc is lower that is a better model for the data used.
What if I then decide to take out one variable because I suspect it doesn't belong and I run 7 explanatory variables with my dependent using OLS and GWR and get an AICc of 5 and 2 respectively. I know I can compare the new AICc values (5 vs 2) for these models but can I compare the AICc value from the second test to the first? For instance knowing that I had an AICc value of 1 tells me that the 8 variables were better than taking one out.
أكثر...
What if I then decide to take out one variable because I suspect it doesn't belong and I run 7 explanatory variables with my dependent using OLS and GWR and get an AICc of 5 and 2 respectively. I know I can compare the new AICc values (5 vs 2) for these models but can I compare the AICc value from the second test to the first? For instance knowing that I had an AICc value of 1 tells me that the 8 variables were better than taking one out.
أكثر...