Bug: Cressie variogram estimator doubles variance

[MuellerSeb]

I found a bug in the cressie variogram estimator. It seams, that the estimated sill is twice as much as the input:

import numpy as np
import gstools as gs

x = np.random.RandomState(19970221).rand(1000) * 100.0
y = np.random.RandomState(20011012).rand(1000) * 100.0
model = gs.Exponential(dim=2, var=2, len_scale=8)
srf = gs.SRF(model, mean=0, seed=19970221)
field = srf((x, y))

bins = np.arange(40)
bin_center, gamma = gs.vario_estimate_unstructured((x, y), field, bins, estimator="cressie")

fit_model = gs.Stable(dim=2)
fit_model.fit_variogram(bin_center, gamma, nugget=False)

print(fit_model)

gives:

Stable(dim=2, var=3.9951222666072255, len_scale=8.859244705331834, nugget=0.0, anis=[1.], angles=[0.], alpha=1.065254356901576)

So it seams, there is a division by 2 missing here:

GSTools/gstools/variogram/estimator.pyx

Line 80 in a01da8e

(1./counts[i] * variogram[i])**4 /

[MuellerSeb]

Fixed here: 0649e61

[MuellerSeb]

And here: cbcef65

[MuellerSeb]

Commits were cherry-picked into #75