Hi all, Paul asked a very good question regarding "Who are the users?" for this type of software.
First, as I explained in my paper, the current trend is to develop software that are tailored to specific applications. As the range of applications of geostatistics widens, it's not feasible to have a piece of software that can handle all the different possible scenarios. Some users manipulate small datasets in 3D. Others analyzes huge datasets (e.g. imagery) but it's restricted to 2D. For others, the temporal dimension is critical... In terms of visualization, some users want 3D exploration and don't care about overlay of GIS layers... For the last 7 years, I have ventured outside the academic bubble and worked for a software company and as a consultant. When you work outside academia, forget about R, Matlab, and so on... Most consultants won't go beyond Excel spreadsheets! I am attending conferences where I have to sell kriging over inverse square distance, so forget about Bayesian kriging, copulas, sequential simulation and all other fancy methods... Definitively in the US, there is a clear disconnect between what's happening in academia and in the real world (maybe less obvious in the Netherlands!). Software companies that want to make a profit won't develop tools that cannot be sold at a large scale, which means accessible to "naive" users.... In terms of GIS, Arcview has almost a quasi monopole for the moment and is basically killing any competition, which is not good news for users who don't like the geostat analyst. As I am finalizing the beta version of our new geostat module, I am struggling with the same problems, how to make geostat accessible to common users.... I agree with the comment about visualization.. Once again, progresses have been accomplished for 3D visualization and the continuing development of google earth should bring more realistic mapping tools soon (I just got a project funded in that arena). regarding your specific comment regarding h-scatterplot crash, our geostat module allows using only a specific percentage of the data for variogram computation since results typically don't change much is you use 10,000 or 100,000 data for example.. Regards, Pierre On Fri, Jan 8, 2010 at 5:19 AM, Paul Hiemstra <p.hiems...@geo.uu.nl> wrote: > Younes Fadakar wrote: > >> Hi all, >> >> I deeply appreciate your invaluable attention answering to my (others! see >> below) questions. I read them carefully. >> Here I should add something to our discussion: >> >> As a user of Matlab, that is the truth, the automation of many of >> processes, say, data preparing, is so easy by means of coding a few lines >> instead of hundreds clicks on buttons, controlling computation core and/or >> graphing features. One wants to do the same task by using S, R, Python or >> SAS. Of course, Matlab is the most powerful for doing this. >> > Hi Younes, > > why is Matlab 'ofcourse' better (I feel the beginning of a flame war :)). > > They are some free similar tools of it e.g FreeMat and Octave. But all of >> them have a weak not interactive graph output! In addition, drawing of >> mega-data, say, as a h-scatter plot is too slow! Sometimes with crashes! >> >> Nowadays, visualization is a subject of sciences rather art. No impressive >> output, no success, no future! I think the developing such a powerful-inside >> impressive-outside tool for geostatistics is not too hard, but a necessary >> for close future. >> >> My list was short but I've tested more software for above two mentioned >> features. >> Handling large dataset is so difficult with all available tools. (If you >> wish you can try!) >> Why? Is this going to be nth mystery of life?! ;) >> Can be this year a new year for Geostatistics paying more attention to >> users? >> > Depends on who you call users. For R+gstat for example I think a lot of > people are very happy with the software, not in the least because there is a > whole community that pays attention to the needs and questions of the users > (wiki, mailing list, etc). > > cheers, > Paul > > Will Geostatistics celebrate its modern fresh look and young heart (core) >> at the end of year? >> I am hopeful. >> >> Regards, >> >> Younes >> >> Beginning of Log >> (Shortened)========================================================= >> << Guillaume wrote: >> "...Vesper. Not very powerful, but the interface is not bad.... >> The command line is not a worry for professionals, so Gstat, GSLIB, R or >> Matlab do the job perfectly fine. Novice users tend to use more something >> within a GIS software for their analyses (e.g. ArcGIS Geostatistical >> toolbox, etc.). Probably the best tool I have seen so far is the GSTAT >> interface used within the Idrisi GIS. >> But you are probably right in saying that there are no user-friendly >> programs that are inexpensive and that would suit both novice users and >> professionals." >> >> << Edzer wrote: >> "...looking for this free or cheap, all-capable package with a complete, >> friendly and robust graphical user interface with dynamic graphics. A >> problem is that such a thing is hard and expensive to develop, and unlikely >> to be arise as a side product of a research project. Look at the worlds of >> GIS or image analysis -- there's a lot of high quality things out there for >> free, but the thing you're looking for is very expensive. >> In your list I missed at least: >> 9. ArcGIS + geostatistical analyst >> 10. SGEMS, the new Stanford software after GSLIB >> 11. other packages in R, such as gstat, randomFields, rsaga, and so on. >> 12. ... (I hope others will finish this list!) >> ...I have the impression that I'm not alone when thinking that although >> graphical, interactive exploratory data analysis is a very nice thing to >> have, a solid data analysis should start from the principle of >> reproducability, and therefore as little as possible depend on the >> reproduction of long sequences of mouse clicks. >> Are users of this list aware of other communities and/or mailing lists >> where considerable activity around geostatistics and/or geostatistical >> software takes place?..." >> >> << Seth wrote: >> "I believe that software for statistical analysis of spatial data must >> realize that many analyses now are on GIS data that covers a large area and >> so is a data rich environment. I have the exact opposite problem from many, >> too many data points! The point and click software that can do geostats >> (IDRISI, ARCMAP, SAM) that I'm familiar with either handles only vector data >> and/or smaller data sets. For instance, a simple semivariogram with a >> sample of the data I had took all night to calculate in IDRISI and came back >> in error. For simple analyses of large data sets, I've taken to writing my >> own code in fortran 90. I can calculate a semivariogram with 1,000s of >> points that cover half of a US state with a huge maximum search distance and >> many bins in about 30 minutes. Most spatial autocorrelation stats are simple >> to write code to calculate...." >> >> << Pierre wrote: >> "Your list is far from complete. I would recommened you take a look at the >> following paper...that provides an overview and comparison of >> functionalities in a series of geostat software, most of them listed on >> ai-geostat website." >> >> << Seth wrote again: >> "I'm wondering what geostatistical software is best for handling very >> large data sets. With the advent of GIS and remote sensing, having too much >> data is a problem. Sampling of course is useful, but only to a point if a >> large study area is used. I've read other places that among the commercial >> stat packages, SAS is best at handling large data sets. Is this true? >> Also, I've produced my own little routine in IDRISI that can create >> 'random' samples that are clustered by inverse distance, so that short lags >> are preferred. Are there any software packages that can create a random >> sample of points that show a pre-specified clustering pattern in space?" >> >> << JanWMerks wrote: >> "I read with a great deal of interest your emessage about Geostatistics in >> pain. Read what I have found out. Real statistics turned into surreal >> geostatistics under the guidance of Professor Dr Georges Matheron..." >> >> << Sebastiano: >> "In general I agree with the comments reported in the preceding replies. >> Then I would add that the problem, if any one exists, doesn't relies on the >> lack of a good gui but maybe on the lack of a kind of standard and >> internationally accepted set of programming routines directed to >> geostatistical analysis. As a final consideration I think that the world of >> spatial analysis is more complex than in the past and a kind of holistic >> view and is needed. For example I'm thinking to other techniques based on >> statistical learning theory,data mining, etc..." >> >> << Paul wrote: >> Being an R user myself (gstat, automap) I would like to comment a little >> on your problem with command line tools. I think what is most important is >> that each application has its own best tool to use. When a novice user wants >> to quickly make some maps, a GUI would probably be the preferred tool. But >> if you, as in my case, want to interpolate thousands of maps, put them on a >> webservice and allow user to get those maps from the web, a GUI tool such as >> ArcGIS is probably not the best option. R is great for these kinds of large >> analysis. In addition, I'm a Linux user and would not trade my command line >> for any GUI :). Therefore I believe a GUI is not per definition better than >> command line. It's just that people are used to GUI nowadays, making command >> line seem old. I agree that it takes quite some time to learn R and that it >> is not a tool suitable for the casual user. A great combo would be a tool >> that has the flexibility and power of R and the ease of use >> of e.g. ArcGIS.Hope you find a solution that suits your particular needs. >> End of Log ====================================================== >> >> King Regards, >> Younes >> >> >> >> >> __________________________________________________________________________________ >> See what's on at the movies in your area. Find out now: >> http://au.movies.yahoo.com/session-times/ >> >> + >> + To post a message to the list, send it to ai-geost...@jrc.ec.europa.eu >> + To unsubscribe, send email to majordomo@ jrc.ec.europa.eu with no >> subject and "unsubscribe ai-geostats" in the message body. DO NOT SEND >> Subscribe/Unsubscribe requests to the list >> + As a general service to list users, please remember to post a summary of >> any useful responses to your questions. >> + Support to the forum can be found at http://www.ai-geostats.org/ >> >> > > > -- > Drs. Paul Hiemstra > Department of Physical Geography > Faculty of Geosciences > University of Utrecht > Heidelberglaan 2 > P.O. Box 80.115 > 3508 TC Utrecht > Phone: +3130 274 3113 Mon-Tue > Phone: +3130 253 5773 Wed-Fri > http://intamap.geo.uu.nl/~paul <http://intamap.geo.uu.nl/%7Epaul> > > > > > + > + To post a message to the list, send it to ai-geost...@jrc.ec.europa.eu > + To unsubscribe, send email to majordomo@ jrc.ec.europa.eu with no > subject and "unsubscribe ai-geostats" in the message body. DO NOT SEND > Subscribe/Unsubscribe requests to the list > + As a general service to list users, please remember to post a summary of > any useful responses to your questions. > + Support to the forum can be found at http://www.ai-geostats.org/ > -- Pierre Goovaerts Chief Scientist at BioMedware Inc. 3526 W Liberty, Suite 100 Ann Arbor, MI 48103 Voice: (734) 913-1098 (ext. 202) Fax: (734) 913-2201 Courtesy Associate Professor, University of Florida Associate Editor, Mathematical Geosciences Geostatistician, Computer Sciences Corporation President, PGeostat LLC 710 Ridgemont Lane Ann Arbor, MI 48103 Voice: (734) 668-9900 Fax: (734) 668-7788 http://goovaerts.pierre.googlepages.com/
