we assumed the uncertainty was normally distributed, otherwise we assumed a lognormal distribution.
This is because when standard errors become larger,
there is a greater chance of sampling negative values for the variables
(which would not make sense).