Loading ...
Sorry, an error occurred while loading the content.

849Re: GEOSTATS: Nugget Effect

Expand Messages
  • srahman@lgc.com
    Jun 1, 1998
      Bill Thayer wrote:

      >My latest request for input concerns the calculation of the value of the
      >variogram at h = 0. I have a highly skewed data set with very high
      >short scale variability. This has caused problems with obtaining a good
      >estimate of the nugget effect, primarily due to 2-3 high values. The
      >data set I am working with contains the analytical results for 25
      >duplicate samples. (I have used the highest of the dupe data in the
      >variogram calculations). I have used this data (in a spreadsheet) to
      >estimate the variogram value at h = 0. I would like to receive feedback
      >on this approach - Is it a valid approach? What other approaches can I

      (a) For highly skewed data, variograms become extremely sensitive
      to those high (possibly outlier?) values. Variability might be dependent
      on location as well (proportional effect). Consider: (1) applying a
      to the data to smooth out high values, e.g. lognormal, but be wary of
      the pros and cons of simply applying a backtransform on kriged or simulated
      values to get the original raw data (there were some exchanges on this
      subject a couple of weeks ago that you might want to skim through in
      the ai-geostats archives); or (2) applying some sort of "robust"
      e.g. the family of general or pairwise relative variograms or nonergodic
      covariance measures to filter out the spatial dependence of variability.

      (b) Out of curiousity, why use the highest value among the duplicates?
      Why not the mean? Or the mode? Or the median?



      *To post a message to the list, send it to ai-geostats@....
      *As a general service to list users, please remember to post a summary
      of any useful responses to your questions.
      *To unsubscribe, send email to majordomo@... with no subject and
      "unsubscribe ai-geostats" in the message body.
      DO NOT SEND Subscribe/Unsubscribe requests to the list!
    • Show all 2 messages in this topic