[R-sig-Geo] JSS Special Issue on Bayesian Statistics
The Journal of Statistical Software (JSS) now reached an impact factor of over 22, and recently ended highest in the category "statistics and probability" on the ISI web of knowledge ranking. Special issues of JSS are very popular and get a lot of citations. We now invite papers for a special issue of the Journal of Statistical Software on "Bayesian Statistics". The conditions are the regular ones of JSS submissions [1], but the special issue aims at bringing together articles presenting open-source software developed by the authors focusing on (but not limited to) the following topics: * Bayesian modelling and inference. * Software for Bayesian Data Analysis. * Software for Bayesian Statistics. * Model selection and assessment. * Priors in Bayesian analysis. * Visualization of results. * Bayesian data analysis of highly structured data. * High-performance computing for Bayesian data analysis. * Bayesian methods for the analysis of ‘Big Data’. Authors who intend to submit an article for this special issue are strongly encouraged to submit paper title, authors, and a preliminary abstract to virgilio.go...@uclm.es or michela.camele...@unibg.it before December 31st, 2018. Guest editors for this special issue are: Dr. Michela Cameletti (University of Bergamo, Italy) Dr. Virgilio Gomez-Rubio (University of Castilla-La Mancha, Spain) Prof. Martyn Plummer (Warwick University, UK) Deadline for papers is Jun 30th, 2019. Submissions must clearly state that they are to be considered for this Special Issue, by mentioning it in the cover letter or by leaving a comment in the upload form. [1] http://www.jstatsoft.org/ With best regards, --- Virgilio Gómez Rubio Escuela de Ingenieros Industriales Universidad de Castilla-La Mancha Avda España s/n 02071 Albacete (Spain) ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] SPDE book for Bayesian spatio-temporal modeling
Dear all, I am happy to announce the forthcoming book “Advanced Spatial Modeling with Stochastic Partial Differential Equations Using R and INLA”. We have a web page at http://www.r-inla.org/spde-book with more information, R code and datasets, and a online (free) Gitbook version. We hope that this will be a useful resource to those of you interested in Bayesian spatial and spatio-temporal modeling. We’d like to thank CRC for agreeing to have a free version of the book on-line. Best, Virgilio [[alternative HTML version deleted]] ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Worked out example of spatiotemporal interpolation of daily precipitation?
Hi, > Blangiardo and Cameletti (2015) Spatial and Spatio-temporal Bayesian > Models with R - INLA models the Parana state rainfall data (chapter > 8). See https://sites.google.com/a/r-inla.org/stbook/ Not kriging but > maybe useful for you. > This is also modeled in the SPDE tutorial http://www.r-inla.org/examples/tutorials/spde-tutorial . Also, I have fitted a space-time model using INLA for the Ireland wind data described in the spacetime vignette. Not really what you want, but you can change the likelihood to a zero-inflated one (INLA provides several of them) if that is what you want… Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Basic questions about Bayesian Spatio-temporal Analysis-INLA
Hi, I think that these model is covered in the book by Michela and Marta. You can also check the book by Banerjee et al. on Bayesian spatial models. I think that this will give a better idea of the different models that you could use. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Basic questions about Bayesian Spatio-temporal Analysis-INLA
Hi Claire, Not sure what type of model or data you are trying to fit. If you have raster data, it would make sense to use them as covariates and not as priors. If you definitely want to fit a spatio-temporal model with INLA you should check this book: http://eu.wiley.com/WileyCDA/WileyTitle/productCd-1118326555.html Also, please check these course materials that I prepared for the GEOSTAT 2017 summer school about spatial model fitting with INLA: https://www.dropbox.com/s/lb9f7eagmmzou5k/materials.zip?dl=0 In a nutshell, the inla() function works similarly as the glm() or gam() functions: you define your model in a formula (which may include random effects) and use a data.frame to pass the data. Hope this helps. Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] simple online CSR pattern generator?
Hi, You can use r-fiddle: http://www.r-fiddle.org/#/fiddle?id=kuUG5uW6 Click on run code and you will see a very simple example. Best, Virgilio [[alternative HTML version deleted]] ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] RFC SEjags: Bayesian spatial econometrics models with JAGS
Hi, I am developing a new package on Bayesian spatial econometrics models that implements the main models using JAGS. Right now, it fits SEM, SLM, SDM, SDEM, SLX and SAC models (using the terminology in the works by LeSage and Page). It also supports a model with a proper CAR distribution for the error term. In addition, probit and logit links (for binary data) are also supported. Computation of the impacts is available through the impacts() function. You can install it from github and run some examples on the Columbus dataset with: devtools::install_github("becarioprecario/SEjags") library(SEjags) example(SEjags) Other models could be easily implemented and the bugs models included in the package can serve as a starting point. I believe that this implementation will be slow for medium or large datasets because of the way it handles inversion of matrices. I’ve noticed that some people on the list work with these models and I would love to hear from them. If you are able to try and have some comments, please, do not hesitate to contact me off list. Best wishes, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] sarprobit question
Dear Jorge, > > I will take a look at the commands to understand why is creating this > W=103x103 matrix. Do you have some suggestions? > You probably have 103 points instead of 102 for some reason… That is what I would check first. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] sarprobit question
Hi, > Thank you very much for your answer. Effectively, your observation is > correct, after I drop a column the command works. > Please, note that dropping the column will give you a 102x102 but, as Obi-Wan Kenobi would say, "this is not the matrix you are looking for”. :) What I mean is that you probably will get a wrong adjacency matrix. :) > I will take a look at the commands to understand why is creating this > W=103x103 matrix. Do you have some suggestions? If you provide a reproducible example (i.e., code plus data) I could look into it. Feel free to contact me off list. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] sarprobit question
Hi, > I first thought that this p=103 was the error, however I did the following: > >wnew <-W[-1,] You need to remove one row AND one column to have a 102x102 matrix. In the code above you are just removing one row. I believe that the error is there. But you should check why you W is 103x103 if you only have 102 data points… Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] R has been accepted for GSoC 2016
Dear all, I know that this is a bit off topic but the R project has been accepted as a mentoring organization for Google Summer of Code 2016. As stated by Gergely in his blog [1]: "In short, Google offers $5,000 to the accepted students to work on open-source and useful R packages for a few months with the help of mentors, who get contributed code instead of money from this nice, alternative and open summer ~internship.” Now it it s the time to develop project proposals that should be included here: https://github.com/rstats-gsoc/gsoc2016/wiki/table-of-proposed-coding-projects Projects can be proposed by students (looking for mentors) and mentors (who will be looking for student). Please, let me encourage you to submit proposals about any spatial project you have in mind. Best wishes, Virgilio P.S: Mentors will also get a very nice t-shirt. :D [1] http://blog.rapporter.net/2016/03/r-projects-at-google-summer-of-code-2016.html ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Prepping data for FlexScan
Dear Paul, > Here is the link to the user manual; there is an image of the matrix > definition file on page 9. > > https://sites.google.com/site/flexscansoftware/download_e/FleXScan%20User%20Guide_e31.pdf?attredirects=0&d=1 > It seems that all the files have the same format: ASCII file with values separated by a space. You can easily take a data.frame and get this type of output with write.table(). But you will need to feed it with the right variables, of course. Also, you may want to check packages DCluster (on CRAN) and dclusterm (on r-forge). dclusterm is a model-based approach to cluster detection and will detect cluster after accounting for different types of covariates (quantitative and qualitative). FlexScan seems to only cope with categorical variables. We are working on dclusterm at the moment and we hope to make an official release on CRAN soon. > Martin Kulldorff is on at least one of the publications with the Japanese > group that designed FlexScan, which is a sign in its favor. He's never > incorporated it into SatScan, however. I think that Barry’s red flag is about Flexscan not being open-source… In any case, this is probably something that could be included in DCluster and/or dclusterm. Not sure if this will be enough for a GSoC project but I will be happy to help. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Excessive whitespace in saved images from sp::plot()
> El 16/12/2015, a las 19:35, Matt Strimas-Mackey > escribió: > > After messing around with parameters aimlessly I managed to solve my > problem. > > The key seems to be that par(mar=c(0, 0, 0, 0)) needs to come AFTER > png('plot.png') > as in: > png('plot.png') > par(mar=c(0, 0, 0, 0)) > plot(square, axes = F, lwd = 2, asp = "", xpd = NA) > dev.off() > > I have no idea why this is the case, but it works! Matt, png() opens a new device with different settings: > par(mar = c(0, 0, 0, 0)) > par()$mar [1] 0 0 0 0 > png(file = "test.png") > par()$mar [1] 5.1 4.1 4.1 2.1 That is why you need to call par() after png(). Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Short course "Spatial and Spatio-Temporal Bayesian Models with R-INLA"
Hi, It seems that the link to the course registration disappeared when I converted the e-mail from HTML to plain text. Apologies for that. The right one is: http://www00.unibg.it/struttura/struttura.asp?cerca=dsaemq_r-inla2015 Best, Virgilio El 13/10/2015, a las 16:17, VIRGILIO GOMEZ RUBIO mailto:virgilio.go...@uclm.es>> escribió: Hi, I believe that this course may be of interest to some of the people following this list. Best, Virgilio --- Short course "Spatial and Spatio-Temporal Bayesian Models with R-INLA" 12-15 January 2016 University of Bergamo (Italy) Instructors Dr. M. Blangiardo - Imperial College London (www.imperial.ac.uk/people/m.blangiardo<http://www.imperial.ac.uk/people/m.blangiardo>) Dr. M. Cameletti - Università di Bergamo (www.unibg.it/pers/?michela.cameletti<http://www.unibg.it/pers/?michela.cameletti>) Dr. V. Gómez Rubio - Universidad de Castilla-La Mancha (www.uclm.es/profesorado/vgomez<http://www.uclm.es/profesorado/vgomez>) Description The course aims at providing an introduction to Bayesian analysis for spatial and spatio-temporal modelling using the R software and R-INLA. The first day of the course will be dedicated to standard spatial analysis with R for different types of data: this includes data import/export, data management and visualisation for geostatistical, area and point pattern data. In the following days the theoretical aspects of the Bayesian approach will be introduced, with a particular focus on spatial and spatio-temporal models and on the Integrated Nested Laplace Approximation (INLA, www.r-inla.org<http://www.r-inla.org>) approach, which has proven to be a valid alternative to the commonly used Markov Chain Monte Carlo simulations. A particular emphasis will be given on examples of applied analysis using the R-INLA package. The course timetable is a mixture of lectures and computer practicals based on the following two books: - Applied spatial data analysis with R (www.asdar-book.org<http://www.asdar-book.org>) - Spatial and spatio-temporal Bayesian models with R-INLA (sites.google.com/a/r-inla.org/stbook<http://sites.google.com/a/r-inla.org/stbook>) Course fee The course fee is 200 Euro for PhD students, 300 Euro for Academia / Public sector and 500 Euro for Private Sector. The fee includes course material, coffee breaks, lunches and the social dinner. For further information and the registration form, please visit the following webpage www.unibg.it/r-inlacourse<http://www.unibg.it/r-inlacourse> Best wishes, Michela Cameletti ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo [[alternative HTML version deleted]] ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] FW: Short course "Spatial and Spatio-Temporal Bayesian Models with R-INLA"
Hi, I believe that this course may be of interest to some of the people following this list. Best, Virgilio --- Short course "Spatial and Spatio-Temporal Bayesian Models with R-INLA" 12-15 January 2016 University of Bergamo (Italy) Instructors Dr. M. Blangiardo - Imperial College London (www.imperial.ac.uk/people/m.blangiardo) Dr. M. Cameletti - Università di Bergamo (www.unibg.it/pers/?michela.cameletti) Dr. V. Gómez Rubio - Universidad de Castilla-La Mancha (www.uclm.es/profesorado/vgomez) Description The course aims at providing an introduction to Bayesian analysis for spatial and spatio-temporal modelling using the R software and R-INLA. The first day of the course will be dedicated to standard spatial analysis with R for different types of data: this includes data import/export, data management and visualisation for geostatistical, area and point pattern data. In the following days the theoretical aspects of the Bayesian approach will be introduced, with a particular focus on spatial and spatio-temporal models and on the Integrated Nested Laplace Approximation (INLA, www.r-inla.org) approach, which has proven to be a valid alternative to the commonly used Markov Chain Monte Carlo simulations. A particular emphasis will be given on examples of applied analysis using the R-INLA package. The course timetable is a mixture of lectures and computer practicals based on the following two books: - Applied spatial data analysis with R (www.asdar-book.org) - Spatial and spatio-temporal Bayesian models with R-INLA (sites.google.com/a/r-inla.org/stbook) Course fee The course fee is 200 Euro for PhD students, 300 Euro for Academia / Public sector and 500 Euro for Private Sector. The fee includes course material, coffee breaks, lunches and the social dinner. For further information and the registration form, please visit the following webpage www.unibg.it/r-inlacourse Best wishes, Michela Cameletti ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] BYM model in R-INLA - how to specify priors individually
Dear James, You should be able to set the prior of your random effects using hyper= in the call to f(). In your particular case, you can try: #Create areas IDs to match the values in nc.adj nc.sids$ID<-1:100 nc.sids$ID2<-1:100 hyper.besag <-list(prec=list(prior="loggamma", params=c(.1, .1))) hyper.iid<-list(prec=list(prior="loggamma", params=c(.001, .001))) #Besag model with random spatial effect (i.e. BYM model) m2<-inla(SID74~NWPROP+ f(nc.sids$ID, model="besag", graph="nc.adj", hyper=hyper.besag)+ f(nc.sids$ID2, model="iid", hyper=hyper.iid), family="poisson", E=nc.sids$EXP, data=as.data.frame(nc.sids), control.predictor=list(compute=TRUE)) Note that I have defined two different indices. You can probably use model="bym" to simplify the formula in your model. Also, run inla.models()$latent$bym to know the default definitions of priors of the hyperparameters. Hope this helps. Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] One-day course on spatial and spatio-temporal modelling with R
Dear R users, Please, see below information about a course on spatial and spatio-temporal modelling with R that we are organising in London. Best wishes, Virgilio -- One-day introductory course on spatial and spatio-temporal modelling with R Faculty of Medicine St. Mary's Campus Imperial College London 9:00-17:00, 16th May 2014 * Course outline: This one-day introductory course is aimed at researchers that have to deal with the analysis of spatial and spatio-temporal data. The course will tackle the problem of analysing spatio-temporal data with the R programming language. Different types of spatial data will be covered, such as point patterns, lattice data and data coming from irregular measurements of continuos processes (geostatistics). In addition, different worked examples will be presented showing how to proceed with the analysis of a wide range of spatial data sets. The topics of the course will contain an introduction to various R packages for the analysis of spatial and spatio-temporal data data. This includes data import/export, data management and visualisation, and how to fit a broad range of models for spatial data. The worked examples will focus on particular real data sets from Epidemiology, Environmental Sciences, Ecology, Economics and others. Although most of the lectures will include live demonstrations of the software, a working knowledge of the R software is desirable to follow the examples. R is distributed as free software, and it can be downloaded from http://cran.r-project.org . Similarly, the course will introduce the statistical concepts behind the analysis, but a basic knowledge of statistics and regression analysis will be necessary. The course will make particular emphasis on Bayesian inference for spatial and spatio-temporal modelling. Hence, examples based on packages R-INLA, BayesX and WinBUGS will be included. We will also discuss the main differences among these and other packages for spatio-temporal modelling. *) Instructor: Dr. Virgilio Gómez-Rubio, Universidad de Castilla-La Mancha *) Organisers: Dr. Marta Blangiardo (Imperial College London) Dr. Virgilio Gómez-Rubio (Universidad de Castilla-La Mancha) *) Booking and course fees: Fees for the course are as follows: - £60 Students - £80 Academics - £160 Others (companies, government related agencies, etc.) Registration for the course can be done on-line at http://www5.imperial.ac.uk/medicine/coursebookings/rmodelling_0514.html Updated information can be found here: http://www.uclm.es/profesorado/vgomez/SSTMR/ ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] DCluster questions
Dear James, > I'm wondering if I can get a little advice on using DCluster. > I have produced a map of areal incidence rates and I'd like to try and detect > clusters. > I have also implemented Bayesian smoothing and have therefore pre and post > smoothing maps. > This seems to work ok however there are a few things I'm confused on: > 1) Should I be using this algorithm on my incidence rates pre or post > Bayesian smoothing ? > I'm thinking that the mle expression above includes a smooth (do I understand > that correctly ?) - > but I' prefer to utilise my hard-won Bayesian smooth if possible. In principle, you should use the observed and expected (from rate standardisation) cases. calculate.mle() just groups the data in a suitable list to be used when resampling from the desired distribution to compute p-values, i.e., this function computes the summary statistics and parameters to be used in the bootstrapping. > 2) Will opgam/kn.iscluster only detect "hotspots" or will they also detect > "coldspots" i.e. areas of statistically unlikely lower incidence rates ? You get a p-value, so you could take the areas with very large p-values as cold-spots. But detection of cold-spots is not of interest, is it? > 3) I'm not familiar with bootstrapping - how many bootstraps should I be > running and why > (i.e. - what should I set R to) ? I believe that 99 should be fine, but you may increase it a bit. With 99 replicates you can find clusters up to a significance level of 0.01. Higher number of replicates will heklp you to detect clusters which are significant below this 0.01 level (which may be useful if you correct for multiple testing, see below). > > 4) How do I decide what the correct value for fractpop is ? I initially had > it set to .25 and I > was getting cluster of 50% of my cases which made no sense. You may try different values of fractpop if you want to detect smaller clusters. > 5) Is there any correction for multiple testing in the opgam() command ? I > have over 3000 areas - > do I need to set a very low alpha ? For the Spatial Scan Statistic you should only pay attention to the significance of the most likely cluster, as this is what Kulldorff's test is testing for. But you can use the p-value reported for the other secondary clusters as a guidance. For GAM, you probably want to make a multiple test correction. Another good option is to plot the centres of the clusters detected as this will give you and idea of the areas with significant high risk. Hope this helps. Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] R2Winbugs - problem with workign directory
Hi James, Try to use the full path (e.g., /home/james/mycode/...) instead of just ".". Also, from the manual page: If 'useWINE=TRUE' is used, all paths (such as 'working.directory' and 'model.file', must be given in native (Unix) style, but 'bugs.directory' can be given in Windows path style (e.g. "c:/Program Files/WinBUGS14/") or native (Unix) style (e.g. "/path/to/wine/folder/dosdevices/c:/Program Files/WinBUGS14"). This is done to achieve greatest portability with default argument value for 'bugs.directory'. Hope this helps. Virgilio El lun, 05-08-2013 a las 07:44 +0100, James Rooney escribió: > Hi all, > > Having a problem with the R2Winbugs working.directory option. > > My call to winbugs looks like this: > > MCMCres<-bugs(debug=T,data=d,inits=inits,parameters.to.save=c("theta","alpha","tau","precv"), > n.chains=2,n.iter=2000,n.burnin=500,n.thin=10, > model.file="ALS_spatial_CAR.txt", program="WinBUGS", WINE=WINE, > WINEPATH=WINEPATH,useWINE=T, > working.directory=".",codaPkg=F) > > If I remove the working.directory option this all works fine. > If I leave it in - winbugs opens but doesn't load any model and doesn't do > anything. It IS finding the model specification file because if it doens't > R2winbugs gives an error. > > I'm totally at a loss as to what is going on. Any ideas ? > > This same line of code runs fine by the way if I open a Windows virtual > machine and run R directly under winbugs - the problem there is the VM has > limited memory and process allocation. > > I hope someone can help! Many thanks, > > James > ___ > R-sig-Geo mailing list > R-sig-Geo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-geo ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] DCluster package
Hi, > I noticed that 'DCluster' package is not available any more while I > self-study 'Applied Spatial Data Analysis with R'. Is there an > alternative? DCluster was archived because of the necessity of complying with recent changes in R internals and a busy (lazy?) maintainer. The package has been fixed and sources are on the main CRAN server now and they will appear in other mirrors soon, as well as compiled versions for Windows and Mac. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Strange results in DCluster package
Dear Corey, Many thanks for reporting this. I have checked your code and the data and I cannot find a problem in kn.iscluster. The only issue is that the log-test statistic in some cases is large (e.g., 12225.064474) and when the test statistic is computed, then R returns Inf. I have changed the code so that argument log.v=TRUE can be set in a call to opgam() and it is passed to kn.iscluster. Hence, the log-test statistic is returned. The package should appear on CRAN in one day or two but I can send you the source code if you are in a hurry. With the new version, you can write: knres<-opgam(data=sa, thegrid=sa[,c("x", "y")], alpha=.05, R=999, iscluster=kn.iscluster, fractpop=.25, model="poisson", mle=mle, log.v=TRUE) and you will get something like this: x ystatistic cluster pvalue size 1 549400.3 3255340 12225.064474 1 0.0011 2 550731.4 3255426 9525.112869 1 0.0014 3 550353.6 3253989 6416.805872 1 0.0018 4 548815.9 3254113 10258.441514 1 0.0013 (...) In order to get the most likely cluster, you can compare the returned (log)-statistics. Hope this helps. Best, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
[R-sig-Geo] Google Summer of Code proposal on spatial epidemiology with R
Dear all, I have submitted a proposal for Google Summer of Code 2011 on spatial epidemiology. A summary of the proposal is shown below, and full information (including other proposals for the R Project) can be found at http://rwiki.sciviews.org/doku.php?id=developers:projects:gsoc2011 Note that GSoC is only open to students. Those interested in spatial epidemiology are encouraged to apply. Please, feel free to contact me off-list if you have any questions. Note that the student application deadline is on the 8th of April (Friday next week). Best wishes, Virgilio == DClusterm: Model-based detection of disease clusters Summary: Model-based detection of disease clusters Background: The analysis of disease data is important in order to detect disease outbreaks and links to risk factors. Some of the methods for cluster detection have been implemented in the DCluster package. However, a model-based approach would be of interest in order to explore disease incidence to potential risk factors. Description: Model-based clustering will be implemented using Generalized Linear Models (in principle, for Poisson and Binomial families). Clustering will be modelled as dummy variables (1=area is in a cluster, 0=area is not is a cluster). Hence, many possible clusters will be proposed and the most likely cluster will be selected according to likelihood ratio test, AIC and (possibly) any other reasonable method. Skills required: : A good working knowledge of R and Generalized Linear Models. Some understanding of spatial statistics will be a plus. Test: Fit a GLM using the North Carolina SIDS data. See example(“readShapePoly”) in maptools package. In this GLM, SID74 will be the outcome and BIR74 a covariate; the Poisson family will be used. In addition, a dummy variable representing a spatial cluster will be included. This dummy variable will include 5 different contiguous regions (i.e., the value of this dummy variable will be 1 for these 5 regions and 0 otherwise). Display the residuals of this model in a map. Mentor: Virgilio Gómez-Rubio, University of Castilla-La Mancha (virgilio.go...@uclm.es) ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Re: [R-sig-Geo] Confidence intervals for non-poisson distributed SMR's
Dear Ben, > I have run into a bit of a problem. In the ASDAR book it shows a method for > calculating Confidence Intervals for Standardised Morbidity Ratios using the > epitools pois.exact function - but my data is overdispersed. The example based on the Poisson-Gamma model will provide credible intervals (because it is a Bayesian model) of the relative risks for over-dispersed data. It is also in the chapter on disease mapping in ASDAR. Hope this helps, Virgilio ___ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo