i am doing some data analysis using a technique like the one pictured here:
in this model, the authors ran a quantile regression to fit those curves to the 95%, 50% and 5% quantiles.
if i were to do the same thing, but fit a lowess curve to the quantiles instead, can i interpret the curves in the same way? ie - according to this fitted model of our observations, approximately 90% of the values would fall between the 5% and 95% curves.
i am less concerned about which approach is more accurate than i am with interpreting it properly.
Advertise on Talk Stats