Open asstergi opened 9 years ago
Thanks for checking out the package.
Line 37: We use a modification to STL which removes the median and the STL seasonal component from the original data. This is how we derive the residual used in the ESD section.
Lines 72-80: This is how we calculate Grubbs' test statistic. We replaced mean and standard deviation with median and median absolute deviation.
Line 51 was done a while back, so we'll go back and look at it again.
@asstergi The 1s were being inserted in order to avoid plotting errors when we log transformed the y axis. However, @ahardjasa submitted a patch that now allows us to support the log transform with 0s in the data.
I updated the code to reflect this, and will submit a patch soon.
Cheers,
Hi guys,
I came across the package which looks great. I have the following 2 questions on the code in 'detect_anoms.R':
1) In line 51, any leading NA's are replaced by 1. Shouldn't it be 0 (zero)?
2) In line 37, the median is subtracted from the data. In lines 72-80, the median is subtracted again. Is this correct?
I don't know the details of 'S-H-ESD' algorithm, so excuse me if I'm wrong!
Thanks!