Remember the long list of assumptions that applied to linear regression? Well, logistic regression is not limited by those same assumptions. However, there are still some important assumptions that we make when using logistic regression:
- Linear relationship with the log odds: As we discussed earlier, underlying logistic regression is an assumption that we can model the log of the odds ratio with a line.
- Encoding of dependent variable: When we set up our model earlier, we assumed that we were trying to predict the probability of B, where a probability of 1.0 corresponded to a positive B example. Thus, we need to prepare our data with this type of encoding. This will be demonstrated in the following ...