Implement hillas features usen eigh

[maxnoe]

The values for length slightly disagree (2 %)

I think this version is much more readable then the 1-4 versions

[kosack]

I like it - it definitely is more readable.

[kosack]

Do you know why the length is a bit different? Rounding error? If so, you could just increase the threshold in the test. It's a bit strange, since length is such a low-order moment that it shouldn't be a problem

[maxnoe]

Do you know why the length is a bit different?

Yes, i think it is the question of sample variance vs. variance. So the question of 1/N vs 1/(N - 1). I kann set ddof=0, than it should be the same

[maxnoe]

Yes, with ddof=0 the results are the same as the other methods. However, with ddof=1 the results should be better for small showers.

[codecov]

Codecov Report

Merging #748 into master will decrease coverage by 0.11%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #748      +/-   ##
==========================================
- Coverage   70.72%   70.61%   -0.12%     
==========================================
  Files         198      198              
  Lines       10669    10688      +19     
==========================================
+ Hits         7546     7547       +1     
- Misses       3123     3141      +18

Impacted Files	Coverage Δ
ctapipe/image/hillas.py	97.61% <100%> (+0.22%)	⬆️
ctapipe/image/tests/test_hillas.py	100% <100%> (ø)	⬆️
ctapipe/tools/camdemo.py	58.97% <0%> (-23.08%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4a784e8...9143dbe. Read the comment docs.