Suppose i have some logit model with, say, 7 explanatory variables and i wish to add another variable, x8. x8 is in the interval [1,5] and i would suspect that low values have a negative effect on the probability of y=1 while a higher value would tend to indicate a higher probability of y=1.
Can there be an advantage to rescaling x8 such that it is, say, defined on [-5,5] ?
I have always thought no, but a colleague recently argued that it would make sense.
Can there be an advantage to rescaling x8 such that it is, say, defined on [-5,5] ?
I have always thought no, but a colleague recently argued that it would make sense.