I am confused about the unit increase on a dataset that has been standardized to have mean = 0 and standard deviation = 1. E.g. the temperature went from 0 to 100 deg initially, and from -1.5 top 1.5 after standardization. What does the unit increase mean in coefficient analysis for logistic regression now? Is this unit still one unit in the non-standardized data, such as temperature goes from 99 to 100, or is it some relative measure, such as temperature goes from -1.5 to -0.5?
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
The unit is now a one standard deviation change. If you have variables with a known unit, like temperature, you are better of not standardizing, as you make the unit harder to interpret.
-
1$\begingroup$ To add to this response: if there is a compelling reason to standardize your predictors prior to fitting the model, e.g. penalized likelihood, you can always unstandardize the resulting coefficient estimates by dividing them by the standard deviation of the corresponding predictor. Then your estimated coefficients can be interpreted on the natural scale. $\endgroup$ Commented Nov 26, 2019 at 18:27
-
$\begingroup$ I see, thank you very much Maarten and psboonstra, I will unstandardize the resulting coefficients in that case! $\endgroup$ Commented Nov 26, 2019 at 19:30