I am running Geographically Weighted Regression (GWR) on a categorical dependent variable so my model is basically a Geographically Weighted Logistic Regression.
I have multiple independent variables, some numerical and some categorical.
While interpreting the results of numerical variables is straight forward, I want to know how to distinguish the reference level of the categorical independent variables and how to interpret those?
let's say I code males as 0 and females as 1. so the coefficients should be interpreted for females as they are coded 1? what if I coded males as1 and females as 2?
Dear Nasser,
There should be no difference in the results if you coded (0,1) or (1,2)
In both cases, the result will be as a ratio between them. The parameter B equals the ratio of males/females (as males variable is coded with the smaller value)
As far as I know, we cannot use 1 and 2 to create dummies in categorical variables, neither for identifying Y-variables in logistic regression, you should use only 0 and 1. Please see reference below:
https://towardsdatascience.com/the-dummys-guide-to-creating-dummy-variables-f21faddb1d40
Regarding the interpretation of the parameters, a binary logistic regression estimates the probability of an event occurring. Your event will be the presence of females, since you have identified them as 1.
If the result of this regression is 0.8, you have an 80% probability of having a female and a 20% probability of being male. Generally, a threshold of 0.5 is stipulated; if greater than 0.5 you have a higher probability of having a female, and if less, a higher probability of having a male. This relationship is inverted if you set 1 for male and 0 for female. Hope this helps. Good luck with your data!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R