I have some gridded data that i am Z transforming, and an inherent part of the data are low frequency but very high/low values. This is absolutely expected in the context of the data. (It is quite normally distributed)

When i calculate the traditional Z-Score (x - mean/sd) and the Modified Z-Score (using spatialEco::raster.Zscore in R), the latter is removing outliers.

The resulting surfaces differ quite markedly; Z-scores for normal method are ca. 80% higher than for modified Z. This pushes many values above 2.

Is this purely about whether i want to keep outliers?

  • $\begingroup$ Getting rid of outliers in an 'automated' way frequently leads to loss of useful information. Usually, delete an observation as an 'outlier' only if you have information from outside the data that it must be wrong or impossible [data entry error (verified against original), equipment failure (noted by tech), 733-year-old patient, negative human weight, etc.] $\endgroup$ – BruceET Jan 29 at 16:40
  • $\begingroup$ Ok thanks, I believe all data should remain even though it really pulls the mean around. $\endgroup$ – Sam Jan 30 at 9:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.