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

AI-GEOSTATS: kriging estimates in S+

Expand Messages
  • vanessa stelzenmüller
    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
    Message 1 of 7 , May 8, 2003
    • 0 Attachment
      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

      --
      * 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.

        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 3 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 4 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 5 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 6 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 7 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.