95% CI

[aminnj]

add a parameter to scale the error band

[aminnj]

Requires generalizing

tidfit/tidfit/fit.py

Line 253 in ed9ce41

sampled_stds = np.nanstd(sampled_ydata, axis=0)

to something like

import numpy as np

sampled_ydata = np.random.normal(0, 1, size=(500, 10))

sampled_ydata[250, 5] = np.nan

nstd = np.nanstd(sampled_ydata, axis=0)
alpha = 0.683  # or 0.95
qstd = np.nanquantile(
    np.abs(sampled_ydata - np.nanmean(sampled_ydata, axis=0)), alpha, axis=0
)
print(nstd)
print(qstd)
# [0.96247188 1.03510003 0.99427915 1.00405951 1.02487994 0.98889515 1.04271745 1.0513755  0.98316079 1.06303003]
# [0.94159716 1.0257386  0.98379726 1.00578981 0.99083196 0.9758496  1.03973236 1.08269368 0.97805165 1.06553879]