The reasoning in the answer by fabee seems correct if applied to the STDE (standard deviation of the error), not the RMSE.
Using similar nomenclature, i=1,…,n is an index representing each record of data, xi is the true value and x^i is a measurement or prediction.
The error ϵi, BIAS, MSE (mean squared error) and RMSE are given by:
ϵi=x^i−xi,BIAS=ϵ¯=1n∑i=1nϵi,MSE=ϵ2¯¯¯¯=1n∑i=1nϵ2i,RMSE=MSE−−−−√.
Agreeing on these definitions, the BIAS corresponds to the sample mean of ϵ, but MSE is not the biased sample variance. Instead:
STDE2=(ϵ−ϵ¯)2¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯=1n∑i=1n(ϵi−ϵ¯)2,
or, if both BIAS and RMSE were computed,
STDE2=(ϵ−ϵ¯)2¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯=ϵ2¯¯¯¯−ϵ¯2=RMSE2−BIAS2.
Note that the
biased sample variance is being used instead of the
unbiased, to keep consistency with the previous definitions given for the MSE and RMSE.
Thus, in my opinion the confidence intervals established by fabee refer to the sample standard deviation of ϵ, STDE. Similarly, confidence intervals may be established for the BIAS based on the z-score (or t-score if n<30) and STDE/n−−√.