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

Re: AI-GEOSTATS: kriging estimates in S+

Expand Messages
  • Isobel Clark
    Vanessa I have no experience with S+ but I would guess it is because the semi-variogram (or covariance) is using the value of the nugget effect at zero
    Message 1 of 7 , May 8, 2003
    • 0 Attachment
      Vanessa

      I have no experience with S+ but I would guess it is
      because the semi-variogram (or covariance) is using
      the value of the nugget effect at zero distance.

      This will tell the kriging system not to honour the
      data values.

      Isobel
      http://geoecosse.bizland.com/whatsnew.htm



      --- vanessa stelzenmüller <vstelzenmueller@...>
      wrote: > Dear list members,
      >
      > I have a question concerning the kriging results
      > derived from "kige" of the spatial modul for S+.
      >
      > The values for the variable of interest estimated
      > with
      > Ordinary Kriging and Universal Kriging at the
      > sampled
      > locations differ from the one observed!
      >
      > Does anyone has a suggestion why?
      >
      >
      > Best wishes
      > Vanessa
      >
      >
      __________________________________________________________________
      >
      > Gesendet von Yahoo! Mail - http://mail.yahoo.de
      > Logos und Klingeltöne fürs Handy bei
      > http://sms.yahoo.de


      __________________________________________________
      Yahoo! Plus
      For a better Internet experience
      http://www.yahoo.co.uk/btoffer

      --
      * To post a message to the list, send it to ai-geostats@...
      * As a general service to the users, please remember to post a summary of any useful responses to your questions.
      * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
      * Support to the list is provided at http://www.ai-geostats.org
    • Isobel Clark
      Vanessa I have no experience with S+ but I would guess it is because the semi-variogram (or covariance) is using the value of the nugget effect at zero
      Message 2 of 7 , May 8, 2003
      • 0 Attachment
        Vanessa

        I have no experience with S+ but I would guess it is
        because the semi-variogram (or covariance) is using
        the value of the nugget effect at zero distance.

        This will tell the kriging system not to honour the
        data values.

        You can test this by adding a short range spherical
        component in instead of the nugget effect. If the
        results change, this is the reason.

        Isobel
        http://geoecosse.bizland.com/whatsnew.htm



        --- vanessa stelzenmüller <vstelzenmueller@...>
        wrote: > Dear list members,
        >
        > I have a question concerning the kriging results
        > derived from "kige" of the spatial modul for S+.
        >
        > The values for the variable of interest estimated
        > with
        > Ordinary Kriging and Universal Kriging at the
        > sampled
        > locations differ from the one observed!
        >
        > Does anyone has a suggestion why?
        >
        >
        > Best wishes
        > Vanessa
        >
        >
        __________________________________________________________________
        >
        > Gesendet von Yahoo! Mail - http://mail.yahoo.de
        > Logos und Klingeltöne fürs Handy bei
        > http://sms.yahoo.de


        __________________________________________________
        Yahoo! Plus
        For a better Internet experience
        http://www.yahoo.co.uk/btoffer

        --
        * To post a message to the list, send it to ai-geostats@...
        * As a general service to the users, please remember to post a summary of any useful responses to your questions.
        * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
        * Support to the list is provided at http://www.ai-geostats.org
      • Edzer J. Pebesma
        The spatial library in S-Plus (or R) does not have a function called kige , or krige . Could you please be more precise what you did? -- Edzer ... -- * To
        Message 3 of 7 , May 9, 2003
        • 0 Attachment
          The spatial library in S-Plus (or R) does not have a function
          called "kige", or "krige". Could you please be more precise
          what you did?
          --
          Edzer

          vanessa stelzenmüller wrote:
          >
          > Dear list members,
          >
          > I have a question concerning the kriging results
          > derived from "kige" of the spatial modul for S+.
          >
          > The values for the variable of interest estimated with
          > Ordinary Kriging and Universal Kriging at the sampled
          > locations differ from the one observed!
          >
          > Does anyone has a suggestion why?
          >

          --
          * To post a message to the list, send it to ai-geostats@...
          * As a general service to the users, please remember to post a summary of any useful responses to your questions.
          * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
          * Support to the list is provided at http://www.ai-geostats.org
        • vanessa stelzenmüller
          Edzer, I used krige under S+ SpatialStats with the specified model (gaussian with nugget) to perform ordinary and universal kriging for two dimensional
          Message 4 of 7 , May 9, 2003
          • 0 Attachment
            Edzer,

            I used "krige" under S+ SpatialStats with the
            specified model (gaussian with nugget) to perform
            ordinary and universal kriging for two dimensional
            spatial data. Afterwards I used "predict" as well as
            "predict.krige" to compute point kriging predictions
            and standard errors at specified locations ( I defined
            a grid). The data are sampled at a regular grid.
            Finally I recognized that the predicted and observed
            values for sampled locations are no identical.

            Best wishes
            Vanessa

            =====
            Vanessa Stelzenmüller,PhD-Student
            ICBM, Dep. Aquatic Ecology
            C.v.O University of Oldenburg
            P.O. Box 2503
            26111 Oldenburg
            <º)))>< <º)))><
            Tel:+49 441-798 3306, Fax:+49 441-798 3701

            __________________________________________________________________

            Gesendet von Yahoo! Mail - http://mail.yahoo.de
            Logos und Klingeltöne fürs Handy bei http://sms.yahoo.de

            --
            * To post a message to the list, send it to ai-geostats@...
            * As a general service to the users, please remember to post a summary of any useful responses to your questions.
            * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
            * Support to the list is provided at http://www.ai-geostats.org
          • Edzer J. Pebesma
            Vanessa, you are right. I was confused, as MASS has a free library called spatial , but you meant the commercial add-on module S+SpatialStats , which is also
            Message 5 of 7 , May 9, 2003
            • 0 Attachment
              Vanessa,

              you are right. I was confused, as MASS has a free library
              called "spatial", but you meant the commercial add-on module
              "S+SpatialStats", which is also called "module spatial" by
              the vendor of S-Plus.

              It seems that S+SpatialStats, when predicting at point locations,
              does as if the locations are shifted a very small amount; I guess
              this is saying the same thing that Isobel meantioned earlier (see
              my sample below). When predicting without a nugget effect, the
              differences are also too large, IMO (see x2 example); I do not
              have an explanation for that.

              I also tried the examples with the (my) gstat S library, which can
              be downloaded from http://www.gstat.org/ , where both examples do
              what you'd expect; again see below.

              Thanks for sharing this with us.
              --
              Edzer

              Here's the output of the S-Plus session:

              S-PLUS : Copyright (c) 1988, 2002 Insightful Corp.
              S : Copyright Lucent Technologies, Inc.
              Version 6.1.2 Release 2 for Linux 2.2.12 : 2002
              Working data will be in .Data
              > module(spatial)
              > x<-as.numeric(1:10)
              > y<-as.numeric(1:10)
              > z<-as.numeric(1:10)
              > d<-data.frame(x=x,y=y,z=z)
              > d
              x y z
              1 1 1 1
              2 2 2 2
              3 3 3 3
              4 4 4 4
              5 5 5 5
              6 6 6 6
              7 7 7 7
              8 8 8 8
              9 9 9 9
              10 10 10 10
              > krige(z~loc(x,y),data=d,covfun=gauss.cov,range=3,sill=1,nugget=1)
              Call:
              krige(formula = z ~ loc(x, y), data = d, covfun = gauss.cov, range = 3, sill =
              1, nugget = 1)

              Coefficients:
              constant
              5.5

              Number of observations: 10
              > x<-krige(z~loc(x,y),data=d,covfun=gauss.cov,range=3,sill=1,nugget=1)
              > predict(x, d)
              x y fit se.fit
              1 1 1 2.827990 1.206120
              2 2 2 2.778034 1.155678
              3 3 3 3.432520 1.152567
              4 4 4 4.299386 1.153139
              5 5 5 5.113279 1.153538
              6 6 6 5.886721 1.153538
              7 7 7 6.700614 1.153139
              8 8 8 7.567480 1.152567
              9 9 9 8.221966 1.155678
              10 10 10 8.172010 1.206120
              > d2<-data.frame(x=y+1e-7,y=y+1e-7)
              > d2
              x y
              1 1 1
              2 2 2
              3 3 3
              4 4 4
              5 5 5
              6 6 6
              7 7 7
              8 8 8
              9 9 9
              10 10 10
              > predict(x, d2)
              x y fit se.fit
              1 1 1 2.827990 1.206120
              2 2 2 2.778034 1.155678
              3 3 3 3.432520 1.152567
              4 4 4 4.299386 1.153139
              5 5 5 5.113279 1.153538
              6 6 6 5.886721 1.153538
              7 7 7 6.700614 1.153139
              8 8 8 7.567480 1.152567
              9 9 9 8.221966 1.155678
              10 10 10 8.172010 1.206120
              > x2<-krige(z~loc(x,y),data=d,covfun=gauss.cov,range=3,sill=2,nugget=0)
              > predict(x2, d)
              x y fit se.fit
              1 1 1 1.000004 0.001272791
              2 2 2 1.999994 0.001272785
              3 3 3 3.000009 0.001272774
              4 4 4 3.999990 0.001272763
              5 5 5 5.000010 0.001272757
              6 6 6 5.999990 0.001272757
              7 7 7 7.000010 0.001272763
              8 8 8 7.999991 0.001272774
              9 9 9 9.000006 0.001272785
              10 10 10 9.999996 0.001272791
              >
              # Now using the gstat library:
              > library(gstat)
              > krige(z~1,~x+y,d,d,vgm(1,"Gau",3,1))
              [using ordinary kriging]
              x y var1.pred var1.var
              1 1 1 1 0.000000e+00
              2 2 2 2 0.000000e+00
              3 3 3 3 5.185705e-33
              4 4 4 4 2.074282e-32
              5 5 5 5 5.185705e-33
              6 6 6 6 2.074282e-32
              7 7 7 7 0.000000e+00
              8 8 8 8 5.185705e-33
              9 9 9 9 0.000000e+00
              10 10 10 10 0.000000e+00
              > krige(z~1,~x+y,d,d,vgm(2,"Gau",3,0))
              [using ordinary kriging]
              x y var1.pred var1.var
              1 1 1 1 7.106437e-33
              2 2 2 2 0.000000e+00
              3 3 3 3 4.440892e-16
              4 4 4 4 2.220446e-16
              5 5 5 5 2.220446e-16
              6 6 6 6 4.440892e-16
              7 7 7 7 2.220446e-16
              8 8 8 8 2.842575e-32
              9 9 9 9 2.842575e-32
              10 10 10 10 0.000000e+00
              >

              --
              * To post a message to the list, send it to ai-geostats@...
              * As a general service to the users, please remember to post a summary of any useful responses to your questions.
              * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
              * Support to the list is provided at http://www.ai-geostats.org
            • Mary C. Christman
              Hello Vanessa, I am guessing that the reason the ordinary krige (OK) and the universal krige (UK) give different results for the observed locations is due to
              Message 6 of 7 , May 9, 2003
              • 0 Attachment
                Hello Vanessa,

                I am guessing that the reason the ordinary krige (OK) and
                the universal krige (UK) give different results for the
                observed locations is due to the trend fit in the
                UK. The results you are getting from the UK are probably
                the predicted values from the trend fit. Does the
                predict.krige function provide the two components (trend
                versus kriged residual) separately or does it give only the
                combined predicted value?

                Mary Christman


                On Fri, 9 May 2003 11:40:32 +0200 (CEST)
                =?iso-8859-1?q?vanessa=20stelzenm=FCller?=
                <vstelzenmueller@...> wrote:

                > Edzer,
                >
                > I used "krige" under S+ SpatialStats with the
                > specified model (gaussian with nugget) to perform
                > ordinary and universal kriging for two dimensional
                > spatial data. Afterwards I used "predict" as well as
                > "predict.krige" to compute point kriging predictions
                > and standard errors at specified locations ( I defined
                > a grid). The data are sampled at a regular grid.
                > Finally I recognized that the predicted and observed
                > values for sampled locations are no identical.
                >
                > Best wishes
                > Vanessa
                >
                > =====
                > Vanessa Stelzenm�ller,PhD-Student
                > ICBM, Dep. Aquatic Ecology
                > C.v.O University of Oldenburg
                > P.O. Box 2503
                > 26111 Oldenburg
                > <�)))>< <�)))><
                > Tel:+49 441-798 3306, Fax:+49 441-798 3701
                >
                > __________________________________________________________________
                >
                > Gesendet von Yahoo! Mail - http://mail.yahoo.de
                > Logos und Klingelt�ne f�rs Handy bei http://sms.yahoo.de
                >
                > --
                > * To post a message to the list, send it to
                > ai-geostats@... * As a general service to the users,
                > please remember to post a summary of any useful responses
                > to your questions. * To unsubscribe, send an email to
                > majordomo@... with no subject and "unsubscribe
                > ai-geostats" followed by "end" on the next line in the
                > message body. DO NOT SEND Subscribe/Unsubscribe requests to
                > the list * Support to the list is provided at
                > http://www.ai-geostats.org

                ----------------------
                Mary C. Christman
                Department of Animal and Avian Sciences
                University of Maryland
                College Park, MD 20742
                fax: 301-405-7980
                office: 301-405-8867
                email: mc276@...


                --
                * To post a message to the list, send it to ai-geostats@...
                * As a general service to the users, please remember to post a summary of any useful responses to your questions.
                * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
                * Support to the list is provided at http://www.ai-geostats.org
              Your message has been successfully submitted and would be delivered to recipients shortly.