Peter EllisJul 30, 2023
I wrote an #rstats blog post to remind myself that just because a log transform improves an ordinary least squares regression in conventional terms (eg makes it more linear and residuals less heteroskedastic), doesn't mean it is more fit for purpose. http://freerangestats.info/blog/2023/07/30/log-transforms
Log transforms, geometric means and estimating population totals

A model that is 'improved' (in terms of making standard assumptions more plausible) by using a logarithm transform of the response will not necessarily be improved for estimating population totals.

free range statistics
Show threadBen BolkerJul 30, 2023
@peter_ellis would doing log transformed analysis plus bias correction (i.e., compute mean of predicted lognormal distribution, exp(mu + sigma^2/2)) give you the best of both worlds?
Show threadPeter Ellis
@bbolker yes
Jul 30, 2023 at 3:06amMastodon for Android
Show threadPeter EllisJul 30, 2023
@bbolker maybe I will do a sequel with that and iterative weighted least squares.
Show threadCameron PatrickJul 30, 2023
@peter_ellis @bbolker (Best of both worlds not guaranteed! AFAIK the standard bias correction is only correct if the underlying population is roughly lognormal; I think this is discussed in Permutt’s paper “Do covariates change the estimand”)