log(A + x), not log(1 + x)

This recent discussion in comments reminds me of something that comes up in regression modeling from time to time. You have a predictor on an outcome which you’d like to model on the log scale—the usual reason, and it’s a … Continue reading → …