AI-GEOSTATS: Moran scatterplot
Hi everybody, I want to ask your opinion on some results from Moran scatterplot. I am working with soil contamination data, and in my opinion the dataset is formed by 2 different distributions, one more diffusive which is the majority of the data, and one generated by a point source process represented by the outliers in the dataset (about 18% found out with box-plot). The dataset is strongly positively skewed. If i use the full dataset and i build the moran scatterplot, i have a global Moran of about 0.02 - no spatial correlation - even though one might expect at least some spatial autocorrelation in the soil contamination data. If i am eliminating all outliers i identified with the box-plot i get a global Moran of about 0.36 - much much better. But if i eliminate only part of the outliers, and not all of them, i get a global moran of 0.49! I would interpret this as: Some outliers (probably the lowest values of my upper outliers - no lower outliers detected by box-plot on my data) belong to the diffusive process, which has a good spatial autocorrelation, while the rest of the data should belong to the point source process. That means not all outliers were generated by the point source pollution, and some are "genuinely" generated by the diffusive process. Since i am dealing with contamination, of course i am interested in what outliers represent, much more if they are above the environmental pollution threshold. Do you think is correct my interpretation? How important is this finding from a statistical point of view? (if it is at all). Thank you so much for your help, Monica -- * To post a message to the list, send it to [EMAIL PROTECTED] * 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 [EMAIL PROTECTED] 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
AI-GEOSTATS: Moran Scatterplot
Hello list, I need your help to interpret this. I am working with contamination data in soil. I think the dataset has two populations, one representing a diffusive process (the majority of the data) and a point source process which generates outliers - or it seems part of them. I used the Moran scatterplot to look at outliers and such and these are my results: If i use the data from one depth layer i have a global Moran of about 0.02 so almost no spatial correlation. If i am eliminating all outliers i identified with the box-plot i get a global Moran of about 0,36 - much, much better. But if i eliminate only part of the outliers, and not all of them, i get a global Moran of 0.49 - extremely good for spatial autocorrelation. I am not sure if i am right but i would interpret this like that: Some outliers (probably thelowest values of my upper outliers - there are no lower outliers - at least detected by box-plot) belong to the diffusive contamination process, which should have a good spatial autocorrelation, while the rest should belong to the point source process. Do you think is it correct my interpretation? How important is this finding in your opinion from a statistical point of view? Thank you so much for any input on that, Monica -- * To post a message to the list, send it to [EMAIL PROTECTED] * 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 [EMAIL PROTECTED] 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
Re: AI-GEOSTATS: Moran scatterplot
Monica The simplest solution to your problem is to use probability paper. If you do not have easy access to this, you can download a free graph paper plotter from http://perso.easynet.fr/~philimar There are also simple algorithms to produce your own. Two populations show up on a probability plot as a line with a definite 'kink' in it. Skewed distributions should be plotted on a logarithmic scale. Explanations can be found in my paper ROKE paper (C&G 1977) which is computer oriented or in my IMGC paper (1993). Both downloadable from http://uk.geocities.com/drisobelclark/resume/Publications.html Isobel Download Yahoo! Messenger now for a chance to win Live At Knebworth DVDs http://www.yahoo.co.uk/robbiewilliams -- * To post a message to the list, send it to [EMAIL PROTECTED] * 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 [EMAIL PROTECTED] 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
RE: AI-GEOSTATS: Moran scatterplot
Title: RE: AI-GEOSTATS: Moran scatterplot -Original Message- From: Pat Bellamy Sent: 28 November 2003 13:57 To: 'Monica Palaseanu-Lovejoy' Subject: RE: AI-GEOSTATS: Moran scatterplot Dear Monica I think it would be worth looking at the following papers as it should give a way of estimating the spatial correlation robustly without having to ignore some of the outliers. Lark, R.M. 2000. A comparison of some robust estimators of the variogram for use in soil survey. European Journal of Soil Science, 51, 137--157. Lark, R.M. 2002. Modelling complex soil properties as contaminated regionalized variables. Geoderma. 106, 171--188. Murray Lark and I have been using these robust estimators on soil contamination data. Yours Pat Mrs Pat Bellamy B.Sc. M.Sc. Statistician/Computer Analyst National Soil Resources Institute (NSRI), Cranfield University at Silsoe, Silsoe,Bedfordshire, MK45 4DT, UK --- Tel.: +44.(0)1525.863260 (direct) Fax.: +44.(0)1525.863253 Email: [EMAIL PROTECTED] http://www.cranfield.ac.uk/nsri/ Conference: Controlling the loss of soil to water To be held at The Geological Society, London on Jan 21st 2004 For more information please visit: <http://www.silsoe.cranfield.ac.uk/nsri/conference/bgrg> Conference: The role of vegetation in environmental protection: Theory and best practice To be held at NSRI, Silsoe between March 24-26th 2004. For more information please visit: http://www.silsoe.cranfield.ac.uk/rpm_conference/ The NSRI is a new Institute that provides research, teaching, training and consultancy in all aspects of the sustainable management of land and soil resources both in the UK and internationally -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Monica Palaseanu-Lovejoy Sent: 28 November 2003 11:43 To: [EMAIL PROTECTED] Subject: AI-GEOSTATS: Moran scatterplot Hi everybody, I want to ask your opinion on some results from Moran scatterplot. I am working with soil contamination data, and in my opinion the dataset is formed by 2 different distributions, one more diffusive which is the majority of the data, and one generated by a point source process represented by the outliers in the dataset (about 18% found out with box-plot). The dataset is strongly positively skewed. If i use the full dataset and i build the moran scatterplot, i have a global Moran of about 0.02 - no spatial correlation - even though one might expect at least some spatial autocorrelation in the soil contamination data. If i am eliminating all outliers i identified with the box-plot i get a global Moran of about 0.36 - much much better. But if i eliminate only part of the outliers, and not all of them, i get a global moran of 0.49! I would interpret this as: Some outliers (probably the lowest values of my upper outliers - no lower outliers detected by box-plot on my data) belong to the diffusive process, which has a good spatial autocorrelation, while the rest of the data should belong to the point source process. That means not all outliers were generated by the point source pollution, and some are "genuinely" generated by the diffusive process. Since i am dealing with contamination, of course i am interested in what outliers represent, much more if they are above the environmental pollution threshold. Do you think is correct my interpretation? How important is this finding from a statistical point of view? (if it is at all). Thank you so much for your help, Monica -- * To post a message to the list, send it to [EMAIL PROTECTED] * 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 [EMAIL PROTECTED] 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 -- * To post a message to the list, send it to [EMAIL PROTECTED] * 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 [EMAIL PROTECTED] 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