http://blog.gmane.org/gmane.comp.lang.r.epi hourly 1 1901-01-01T00:00+00:00 http://gmane.org/img/gmane-25t.png http://gmane.org http://comments.gmane.org/gmane.comp.lang.r.epi/150 <pre> I would like to generated a nested case-control set, using multiple factors to match. I am using the ‚ccwc‘ function, and while I apparently get the right behavior, I end up with a combo of misnamed and unnamed columns. I am unsure whether this is a bug, or whether I am making an inappropriate call. Example: ----- suppressMessages(library(Epi)) data(diet) # Example from the Documentation suppressWarnings(head(ccwc(doe, dox, chd, origin = dob, controls = 2, data = diet, include = energy, match = job))) ## ## Sampling risk sets: ............................................. ## Set Map Time Fail job energy ## 1 1 3 1984-03-20 1 Conductor 24.95 ## 2 1 230 1982-03-14 0 Conductor 36.18 ## 3 1 46 1978-10-08 0 Conductor 22.37 ## 4 2 9 1976-07-27 1 Conductor 23.16 ## 5 2 166 1981-12-09 0 Conductor 27.03 ## 6 2 253 1973-05-13 0 Conductor 32.71 # Exactly as it should, given this is the example from the documentation # Then..., trying matching on multiple factors. # Matching on multiple factors results in mislabeled columns, as well as # unlabeled columns suppressWarnings(head(ccwc(doe, dox, chd, origin = dob, controls = 2, data = diet, include = list(height, weight, fat), match = list(job, energy.grp)))) ## ## Sampling risk sets: .............................................. ## Set Map Time Fail height weight fat NA. NA..1 ## 1 1 3 1984-03-20 1 Conductor &lt;=2750 KCals 152.4 49.90 11.249 ## 2 1 70 1978-06-01 0 Conductor &lt;=2750 KCals 170.2 67.59 10.702 ## 3 1 149 1983-11-03 0 Conductor &lt;=2750 KCals 170.0 59.51 10.798 ## 4 2 9 1976-07-27 1 Conductor &lt;=2750 KCals 174.0 66.00 10.140 ## 5 2 187 1972-04-05 0 Conductor &lt;=2750 KCals 172.7 71.90 10.019 ## 6 2 13 1972-03-23 0 Conductor &lt;=2750 KCals 174.0 61.46 7.272 ----- I am ending up with the the retained columns at the end, as well as the columns used for matching. However, the names for the included columns are not assigned properly, and there while the names for the matched columns are not retained. Is this a bug, or am I making an inappropriate call to the ‚ccwc‘ function? Thanks, -albert Þessi tölvupóstur og viðhengi gæti innihaldið trúnaðarupplýsingar og/eða einkamál og er eingöngu ætlaður þeim sem hann er stílaður á. Efni hans og innihald er á ábyrgð þess starfsmanns sem sendir hann ef það tengist ekki starfsemi Hjartaverndar. Ef sending þessi hefur ranglega borist yður vinsamlega gætið fyllsta trúnaðar,tilkynnið sendanda og eyðileggið sendinguna eins og skylt er skv. 5. mgr. 47. gr. laga um fjarskipti nr. 81/2003 um fjarskipti. The information transmitted, including any attachment, may contain confidential and/or privileged material and is intended only for the addressee only. The contents of the message are the individual senders responsibility if it is not related to the operation of Hjartavernd. If you receive this in error, please keep the information confidential, contact the sender and delete the material from your system. _______________________________________________ R-sig-Epi&lt; at &gt;r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-epi </pre> Albert V. Smith 2013-05-07T13:11:40 http://comments.gmane.org/gmane.comp.lang.r.epi/149 <pre>Hi there, I've been having a few issues. My data set is as follows str(new.data)'data.frame':345 obs. of 11 variables: $ WSZ_Code : int 2 6 7 7 7 5 1 5 8 1 ... $ Treatment_Code: int 3 1 4 4 4 2 2 2 1 2 ... $ Year : int 1996 1996 1996 1996 1996 1996 1996 1996 1996 1996 ... $ Month : int 1 2 2 2 3 3 3 3 3 3 ... $ TTHM : num 30.7 24.8 60.4 58.1 62.2 40.3 20.8 36.3 40.5 47.8 ... $ CL2_FREE : num 0.35 0.25 0.05 0.15 0.2 0.15 0.15 0.025 0.25 0.05 ... $ BrO3 : num 0.5 0.5 0.5 0.5 0.5 ... $ Colour : num 0.75 0.75 0.75 0.75 2 2 0.75 1.9 1.9 1.9 ... $ PH : num 7.4 6.9 7.1 7.5 7.6 7.7 6.8 7.9 7.4 8.2 ... $ TURB : num 0.055 0.2 0.055 0.055 0.055 0.055 0.11 0.11 0.055 0.16 ... $ seasons : Factor w/ 4 levels "autumn","spring",..: 4 4 4 4 2 2 2 2 2 2 ... The aim is to try to find exposure estimates for TTHM. I try the heirarchical mixed effects models such as:mod3 &lt;- lme(tthm ~ cl2free, random= ~ 1| treatcode/loc_code, data=new.data, method ="ML") #loc_code=WSZ_Code which seem to work but whenever I try eg: plot(augPred(mod3)) I get this error:Error in plot(augPred(mod3)) : error in evaluating the argument 'x' in selecting a method for function 'plot': Error in sprintf(gettext(fmt, domain = domain), ...) : invalid type of argument[1]: 'symbol' I think maybe it's because all the integers in the dataset are actually factors. Any ideas how to change this and find a solution? Thanks [[alternative HTML version deleted]] </pre> Omnia Abdulrazeg 2013-05-02T15:41:38 http://comments.gmane.org/gmane.comp.lang.r.epi/148 <pre>R epi friends, Im conducting some initial longitudinal analysis of some biomarkers. I would like to have some idea of how the crude trajectories behave before the attempt to run marginal models. I looked around and found the stats::interaction.plot and epicalc::followup.plot. However, due to a unknown reason, these functions were truncating the trajectories before the maximum time. After a while banging my head around I decided to make my own function, which I share with you below. It is very similar to followup.plot, but in an attempt to make a less polluted plot, instead of selecting randomly fewer subject it selects n subjects closest to the initial median value, initial lower value and initial maximum value. In hope it may be useful to others too... Kind regards, # data = dataset in long format # id = character indicating the name of an unique id variable for each patient in data (tested with numeric variable) # time = a character indicating the name of a numeric variable representing the time of the measure # outcome = a character indicating the name of a numeric variable representing the outcome # plot.arg = a list of arguments to pass to plot.default # line.arg = one or more lists of arguments to pass to lines (one list to each strata of by) # zap = if auto then the n subjects with the lower, median, higher initial outcome values will be ploted, anything different from "auto" will plot all subjects # n = number of subject to use in zap # by = a character indicating the name of a factor/character variable representing strata for ploting # leg.args is a list of arguments to pass to legend, it will be used only if by is different from NULL # think about making conditions for zap such as only the median, only the lowest or only the highest, the final highest, final lowest, final median etc zapplot &lt;- function(data,id,time,outcome, plot.arg=list(xlab='time',ylab=outcome), line.arg=list(list(lty=2,col='blue'),list(lty=3,col='red')), leg.args=list(x='top',inset=c(0,-.15), horiz=TRUE,xjust=.5,col=c('blue','red'),lty=2:3, bty='n',xpd=T), zap='auto',n=1,by=NULL){ # if(!is.null(by)){ # if(length(line.arg)!=nlevels(by)){ # stop('If by is porvided, number of lists in line.arg and number of levels in by should be the same! Try improving the lists in line.arg.')} # } data &lt;- data[which(!is.na(data[,time]) &amp; !is.na(data[,outcome])),] plot.arg &lt;- c(list(x = data[,time],y = data[,outcome],type = 'n'),plot.arg) do.call(plot,plot.arg) if(!is.null(by)){ if(!is.factor(data[,by])){ data[,by] &lt;- as.factor(data[,by]) } tmp2 &lt;- data for(j in 1:nlevels(data[,by])){ # i=2 data &lt;- tmp2 data &lt;- data[which(data[,by]==levels(data[,by])[j]),] tmp &lt;- data[which(data[,time]==min(data[,time])),] tmp &lt;- tmp[order(tmp[,outcome]),] # tmp[,c(id,time,outcome)] id.max &lt;- tmp[(nrow(tmp)-(n-1)):nrow(tmp),id] # id.min &lt;- tmp[(1:n),id] # data[which(data$id==id.min),outcome] tmp$med &lt;- abs(tmp[,outcome]-median(tmp[,outcome],TRUE)) tmp &lt;- tmp[order(tmp$med),] id.med &lt;- tmp[1:n,id] if(zap=='auto'){ for(i in 1:length(id.min)){ line.arg.min &lt;- c(list(x = data[which(data[,id]==id.min[i]),time],y = data[which(data[,id]==id.min[i]),outcome]),line.arg[[j]]) do.call(lines,line.arg.min) } for(i in 1:length(id.med)){ line.arg.med &lt;- c(list(x = data[which(data[,id]==id.med[i]),time],y = data[which(data[,id]==id.med[i]),outcome]),line.arg[[j]]) do.call(lines,line.arg.med) } for(i in 1:length(id.max)){ line.arg.max &lt;- c(list(x = data[which(data[,id]==id.max[i]),time],y = data[which(data[,id]==id.max[i]),outcome]),line.arg[[j]]) do.call(lines,line.arg.max) } } else{ uni &lt;- unique(data[,id]) for(i in 1:length(uni)){ line.arg.all &lt;- c(list(x=data[which(data[,id]==uni[i]),time],y=data[which(data[,id]==uni[i]),outcome]),line.arg[[j]]) do.call(lines,line.arg.all) } } } # COLOCAR AQUI OS COMANDOS DE LEGENDA para identificar as diferentes categorias do by leg.args &lt;- c(legend=list(levels(data[,by])),leg.args) do.call(legend,leg.args) } else{ tmp &lt;- data[which(data[,time]==min(data[,time])),] tmp &lt;- tmp[order(tmp[,outcome]),] # tmp[,c(id,time,outcome)] id.max &lt;- tmp[(nrow(tmp)-(n-1)):nrow(tmp),id] # id.min &lt;- tmp[(1:n),id] # data[which(data$id==id.min),outcome] tmp$med &lt;- abs(tmp[,outcome]-median(tmp[,outcome],TRUE)) tmp &lt;- tmp[order(tmp$med),] id.med &lt;- tmp[1:n,id] if(zap=='auto'){ for(i in 1:length(id.min)){ line.arg.min &lt;- c(list(x = data[which(data[,id]==id.min[i]),time],y = data[which(data[,id]==id.min[i]),outcome]),line.arg[[1]]) do.call(lines,line.arg.min) } for(i in 1:length(id.med)){ line.arg.med &lt;- c(list(x = data[which(data[,id]==id.med[i]),time],y = data[which(data[,id]==id.med[i]),outcome]),line.arg[[1]]) do.call(lines,line.arg.med) } for(i in 1:length(id.max)){ line.arg.max &lt;- c(list(x = data[which(data[,id]==id.max[i]),time],y = data[which(data[,id]==id.max[i]),outcome]),line.arg[[1]]) do.call(lines,line.arg.max) } } else{ uni &lt;- unique(data[,id]) for(i in 1:length(uni)){ line.arg.all &lt;- c(list(x=data[which(data[,id]==uni[i]),time],y=data[which(data[,id]==uni[i]),outcome]),line.arg[[1]]) do.call(lines,line.arg.all) } } } } # zapplot(b,'pront','dia','HLA',zap='all',n=2,by='CD4cod') # Dr. Pedro Emmanuel A. A. do Brasil Curriculum Lattes: http://lattes.cnpq.br/6597654894290806 Instituto de Pesquisa Clínica Evandro Chagas Fundação Oswaldo Cruz Rio de Janeiro - Brasil Av. Brasil 4365, CEP 21040-360, Tel 55 21 3865-9648 email: pedro.brasil-am19qAsr2QmNrImTOV1V8w&lt; at &gt;public.gmane.org email: emmanuel.brasil-Re5JQEeQqe8AvxtiuMwx3w&lt; at &gt;public.gmane.org ---Apoio aos softwares livres www.zotero.org - gerenciamento de referências bibliográficas. www.broffice.org ou www.libreoffice.org - textos, planilhas ou apresentações. www.epidata.dk - entrada de dados. www.r-project.org - análise de dados. www.ubuntu.com - sistema operacional [[alternative HTML version deleted]] </pre> Pedro Emmanuel Alvarenga Americano do Brasil 2013-04-02T02:44:35 http://comments.gmane.org/gmane.comp.lang.r.epi/147 <pre>Hands-on Webinar (no charge) Advances in Regression: Modern Ensemble and Data Mining Approaches **Part of the series: The Evolution of Regression from Classical Linear Regression to Modern Ensembles Register Now for Parts 3, 4: https://www1.gotomeeting.com/register/500959705 **All registrants will automatically receive access to recordings of Parts 1 &amp; 2. Course Abstract: Overcoming Linear Regression Limitations Regression is one of the most popular modeling methods, but the classical approach has significant problems. This webinar series addresses these problems. Are you working with larger datasets? Is your data challenging? Does your data include missing values, nonlinear relationships, local patterns and interactions? This webinar series is for you! In our March 29th session (Part 3), we will focus on modern ensemble and data mining approaches. These methods dramatically improve the performance of weak learners such as regression trees. The techniques discussed here enhance the performance of regression trees considerably. These methods inherit the good features of trees (variable selection, missing data, mixed predictors) and improve on the weak features such as prediction performance. Did you miss parts 1 and 2? With your registration, you will receive links to the recordings of Part 1 and 2. Covered in part 1 and 2 are improvements to conventional and logistic regression, as well as a discussion of classical, regularized, and nonlinear regression from both a theoretical and hands-on point of view. The hands-on component includes a step-by-step demonstration with instructions for reproducing the demo at your leisure. Especially for the dedicated student: after watching this recording, you will be able to apply these methods to your own data. Part 3: March 29, 10-11am PST - Regression methods discussed: * Nonlinear Ensemble Approaches: o TreeNet Gradient Boosting o Random Forests o Gradient Boosting incorporating Random Forests * Ensemble Post-Processing: o ISLE Importance Sampled Learning Ensembles o RuleLearner rule based learning ensembles Part 4: April 12, 10-11am PST - Hands-on demonstration of concepts discussed in Part 3 * Step-by-step demonstration * Datasets and software available for download * Instructions for reproducing demo at your leisure * For the dedicated student: apply these methods to your own data (optional) [[alternative HTML version deleted]] </pre> Lisa Solomon 2013-03-20T20:04:15 http://comments.gmane.org/gmane.comp.lang.r.epi/146 <pre>Begins Friday (no charge) The Evolution of Regression: An Upcoming Webinar Series (Hands-on Component) Registration: http://bit.ly/salford-systems-regression-webinar-series Regression is one of the most popular modeling methods, but the classical approach has significant problems. This webinar series address these problems. Are you are working with larger datasets? Is your data challenging? Does your data include missing values, nonlinear relationships, local patterns and interactions? This webinar series is for you! We will cover improvements to conventional and logistic regression, and will include a discussion of classical, regularized, and nonlinear regression, as well as modern ensemble and data mining approaches. This series will be of value to any classically trained statistician or modeler. Overcoming Linear Regression Limitations Part 1: March 1 - Regression methods discussed • Classical Regression • Logistic Regression • Regularized Regression: GPS Generalized Path Seeker • Nonlinear Regression: MARS Regression Splines Part 2: March 15 - Hands-on demonstration of concepts discussed in Part 1 • Step-by-step demonstration • Datasets and software available for download • Instructions for reproducing demo at your leisure • For the dedicated student: apply these methods to your own data (optional) Part 3: March 29 - Regression methods discussed *Part 1 is a recommended pre-requisite • Nonlinear Ensemble Approaches: TreeNet Gradient Boosting; Random Forests; Gradient Boosting incorporating RF • Ensemble Post-Processing: ISLE; RuleLearner Part 4: April 12 - Hands-on demonstration of concepts discussed in Part 3 • Step-by-step demonstration • Datasets and software available for download • Instructions for reproducing demo at your leisure • For the dedicated student: apply these methods to your own data (optional) </pre> Lisa Solomon 2013-02-26T21:29:41 http://comments.gmane.org/gmane.comp.lang.r.epi/144 <pre>Hello to all, I have a problem using the Epi package and I would like your help. my data set looks like this: id date of born date of entry date of first screen date of exit the date of entry is fixed for all obs to be a specific date. Now, I want to calculate person years of no screening, thus date first screen-date of entry and person years of screening therefore, date of exit-date of first screen. I have created two lexis objects, one for the non-screening period calculating the age at entry (date of entry-date of birth) , and years of no screening and one for the screening period calculating again the same age and person years of screening. Then I want to create an age/time matrix with person years, so I split the lexis object first on age and then on calendar year with splitLexis function. Finally, I tabulate in a matrix and all seems fine for the person years for the non exposed period. However, for the screening period I got a warning message saying that I have multiple time bands when I am trying to tabulate into the final matrix. this is how my algorithm looks like ######################################## ####calculation of no-screening years### ######################################## library(foreign) data1&lt;-read.dta("T:\\person_no_screening_years.dta") attach(data1) names(data1) bbb&lt;-data.frame(data1) # Convert the character dates into numerical variables (fractional years)# library(Epi) dobt &lt;- cal.yr( bbb$date_born, format="%d/%m/%Y" ) doet &lt;- cal.yr( bbb$date_entry, format="%d/%m/%Y" ) doxt &lt;- cal.yr( bbb$min_scr_date , format="%d/%m/%Y" ) # Set up the dataset as a Lexis object # # with age and calendar time as time-scales # cohh &lt;- Lexis( entry = list( per=doet, age=doet-dobt ), exit = list( per=doxt ), exit.status = cases, data = bbb ) # Split time along one time-axis # La1 &lt;-splitLexis( cohh, "aget", breaks=c(30,35,40,45,50,55,60,65,70,75,80,85,90,95)) # Split time along the calendar time-axes # Lap1 &lt;-splitLexis( La1 , "per", breaks=c(1975,1980,1985,1990,1995,2000,2005,2010)) # Tabulate events and person-years # PYtab2 &lt;- with( Lap1, xtabs( cbind( C=lex.Xst, Y=lex.dur ) ~ timeBand(Lap1,"aget","left") + timeBand(Lap1,"per","left") ) ) names(dimnames(PYtab2)) &lt;- c("","Period","Aget") T&lt;-round(ftable( PYtab2, row.vars=c(3,1) )) T write.table(T,"T:\\Years_no_screening.txt") ################################################################# ################################################################# ###################################### ####calculation of screening years#### ###################################### library(foreign) data2&lt;-read.dta("T:\\person_screening_years_new.dta") attach(data2) names(data2) vvvv&lt;-data.frame(data2) # Convert the character dates into numerical variables (fractional years)# library(Epi) dob &lt;- cal.yr( vvvv$date_born, format="%d/%m/%Y" ) doe &lt;- cal.yr( vvvv$date_entry, format="%d/%m/%Y" ) dox &lt;- cal.yr( vvvv$date_exit , format="%d/%m/%Y" ) doz &lt;- cal.yr( vvvv$min_scr_date, format="%d/%m/%Y" ) # Set up the dataset as a Lexis object # # with age and calendar time as time-scales # cohR &lt;- Lexis( entry = list( per=doz, age1=doe-dob ), exit = list( per=dox ), exit.status = cases, data = vvvv ) # Split time along one time-axis # LaR &lt;-splitLexis( cohR, "age1", breaks=c(30,35,40,45,50,55,60,65,70,75,80,85,90,95) ) # Split time along the calendar time-axes # LapR &lt;-splitLexis( LaR , "per", breaks=c(1975,1980,1985,1990,1995,2000,2005,2010)) # Tabulate events and person-years # PYtab3R &lt;- with( LapR, xtabs( cbind( C=lex.Xst, Y=lex.dur ) ~ timeBand(LapR,"age1","left") + timeBand(LapR,"per","left") ) ) names(dimnames(PYtab3R)) &lt;- c("","Period","Age") FR&lt;-round(ftable( PYtab3R, row.vars=c(3,1) )) FR write.table(FR,"T:\\Years_screening.txt") tapply( status(LapR,"exit")==1, list( timeBand(LapR,"age1","left"), timeBand(LapR,"per","left") ), sum ) tapply( dur(LapR), list( timeBand(LapR,"age1","left"), timeBand(LapR,"per","left") ), sum ) Could you please help me with that??? [[alternative HTML version deleted]] </pre> dimitris m 2013-02-25T15:52:21 http://comments.gmane.org/gmane.comp.lang.r.epi/143 <pre>Course in: STATISTICAL PRACTICE IN EPIDEMIOLOGY USING R IARC, Lyon, Fance, Thursday 23 to Wednesday 29 May 2013. http://BendixCarstensen.com/SPE The course is intended for epidemiologists and statisticians who wish to use R (www.r-project.org) for statistical modelling and analysis of epidemiological data. The aim of the course is to give participants access to a set of tools in current use by statisticians in epidemiology. The course requires fairly good understanding of statistical principles and some familiarity with epidemiological concepts and study types. The course will be mainly practically oriented with more than half the time at the computer. Contents: * History of R. Language. Objects. Functions. * Interface to other data formats. Data frames. * Tabulation of data. * Logistic regression for case-control-studies. * Poisson regression for follow-up studies. * Causal inference. * Parametrization of models. * Graphics in R. * Graphical reporting of results. * Time-splitting &amp; Standardized Mortality Ratios. * Survival analysis and competing risks. * Multi-state models. * Nested and matched case-control studies. * Case-cohort studies. Venue: IARC, Lyon, France Price: 600 EUR The course is a self-supporting entity, and there are NO grants or stipends available for travel or participation. Deadline for application: 1st April 2013. Information to applicants will be given by e-mail shortly after the deadline. Further information and application form at: http://BendixCarstensen.com/SPE ----------------------------------------------------------------------- This message and its attachments are strictly confidenti...{{dropped:8}} </pre> Martyn Plummer 2013-02-22T11:47:39 http://comments.gmane.org/gmane.comp.lang.r.epi/142 <pre>Dear List Member, Firstly, apologies if you've received this request via multiple lists, this is just a brief announcement and request for help re. R and High Performance Computing. Over the past 3 years, the open-source SPRINT project has focussed on the parallelisation of R functions used by biostatisticians and other data analysts. On the basis of feedback from the biostatistical community in 2010, the SPRINT team at the University of Edinburgh developed 7 parallelised functions of generic utility in the analysis of large data matrices. More information can be found here: www.r-sprint.org . The latest version of SPRINT (v1.0.4) is now available from CRAN and includes, for the first time, the ability to run the software on Apple OS X. SPRINT has always been scalable from desktop to cluster to HPC facility, however, it can now take advantage of multi-core hardware in both Linux and OS X environments. SPRINT v1.0.4 software and documentation is available here: http://cran.r-project.org/web/packages/sprint/index.html A request for help… The SPRINT team would like to revise and refresh our understanding of the needs and requirements of the R community for High Performance computing. To achieve this, we've written a brief questionnaire (no more than 15 mins) which we hope will allow us to capture needs and prioritise SPRINT development over the next 18 months. It would be incredibly helpful if you could take a few moments to complete this questionnaire and tell us more about your R/HPC usage and/ or problems you may have with large and demanding data analyses. The questionnaire can be found at the link below and will be open until 4th March. After the questionnaire has closed, we'll analyse the data, make the results available (most likely via r-sprint.org) and prioritise our development of new functionality for SPRINT. https://www.survey.ed.ac.uk/2013_sprint/ Thanks very much in advance for your help with our requirements analysis! Apologies again if you receive a similar request from us via mailing lists etc. to which you subscribe. If you have any queries regarding the use of SPRINT (including installation), please feel free to contact us at: sprint-5WhEfG1TI8k&lt; at &gt;public.gmane.org All the best, Kevin Robertson ------------------------------------------------ SPRINT - A parallel framework for R sprint-5WhEfG1TI8k&lt; at &gt;public.gmane.org www.r-sprint.org </pre> SPRINT 2013-02-18T14:22:20 http://comments.gmane.org/gmane.comp.lang.r.epi/141 <pre>The Epi package have just been updated to 1.1.44, and is now on CRAN, http://cran.r-project.org/web/packages/Epi/index.html - a bug in ROC have been fixed. Now the AUC is computed correctly. - the functionality of boxes.Lexis have been - errors in computing residuals in one of the models in apc.fit now fixed And a number of other smaller updates. Best regards, Bendix Package maintainer, Epi http://BendixCarstensen.com/Epi ______________________________________________ Bendix Carstensen Senior Statistician Epidemiology Steno Diabetes Center A/S Niels Steensens Vej 2-4 DK-2820 Gentofte Denmark +45 44 43 87 38 (direct) +45 30 75 87 38 (mobile) bxc-77mEMlKuB9s&lt; at &gt;public.gmane.org http://BendixCarstensen.com www.steno.dk </pre> BXC (Bendix Carstensen 2013-02-08T08:31:09 http://comments.gmane.org/gmane.comp.lang.r.epi/139 <pre>submitted [[alternative HTML version deleted]] </pre> Beiranvand 2012-12-17T08:25:12 http://comments.gmane.org/gmane.comp.lang.r.epi/138 <pre>Webinar: Advances in Gradient Boosting: the Power of Post-Processing TOMORROW: December 14, 10-11 a.m., PST Webinar Registration: http://2.salford-systems.com/gradientboosting-and-post-processing/ Course Outline: I. Gradient Boosting and Post-Processing: o What is missing from Gradient Boosting? o Why post-processing techniques are used? II. Applications Benefiting from Post-Processing: Examples from a variety of industries. o Financial Services o Biomedical o Environmental o Manufacturing o Adserving III. Typical Post-Processing Steps IV. Techniques: o Generalized Path Seeker (GPS): modern high-speed LASSO-style regularized regression. o Importance Sampled Learning Ensembles (ISLE): identify and reweight the most influential trees. o Rulefit: ISLE on “steroids.” Identify the most influential nodes and rules. V. Case Study Example: o Output/Results without Post-Processing o Output/Results with Post-Processing o Demo</pre> Lisa Solomon 2012-12-13T20:21:19 http://comments.gmane.org/gmane.comp.lang.r.epi/137 <pre>Webinar: Advances in Gradient Boosting: the Power of Post-Processing TOMORROW: December 14, 10-11 a.m., PST Webinar Registration: http://2.salford-systems.com/gradientboosting-and-post-processing/ Course Outline: I. Gradient Boosting and Post-Processing: o What is missing from Gradient Boosting? o Why post-processing techniques are used? II. Applications Benefiting from Post-Processing: Examples from a variety of industries. o Financial Services o Biomedical o Environmental o Manufacturing o Adserving III. Typical Post-Processing Steps IV. Techniques: o Generalized Path Seeker (GPS): modern high-speed LASSO-style regularized regression. o Importance Sampled Learning Ensembles (ISLE): identify and reweight the most influential trees. o Rulefit: ISLE on "steroids." Identify the most influential nodes and rules. V. Case Study Example: o Output/Results without Post-Processing o Output/Results with Post-Processing o Demo [[alternative HTML version deleted]] </pre> Lisa Solomon 2012-12-13T19:16:28 http://comments.gmane.org/gmane.comp.lang.r.epi/136 <pre>Dear all, An updated version, 1.1.40, of Epi is now available on CRAN. A number of small bugs have been fixed; the main amendments are: effx now allows RR for binary data (via the log-link for binomial outcome) and RD for follow-up date (via the identity link for Poisson data). Existing code will work exactly as before. Relevel.Lexis now facilitates merging of states in a multistate Lexis object. b.r. Bendix Carstensen Package maintainer ______________________________________________ Bendix Carstensen Senior Statistician Epidemiology Steno Diabetes Center A/S Niels Steensens Vej 2-4 DK-2820 Gentofte Denmark +45 44 43 87 38 (direct) +45 30 75 87 38 (mobile) bxc-77mEMlKuB9s&lt; at &gt;public.gmane.org http://BendixCarstensen.com www.steno.dk </pre> BXC (Bendix Carstensen 2012-11-06T07:24:33 http://comments.gmane.org/gmane.comp.lang.r.epi/134 <pre>R friends, Before coming to this post I did take a look at Rothman's book and Sklo's book, but not very iluminating so far. Recently I was challenged with a problem of causal diagrams. A partner o mine came with a dataset with more then 1500 observations and more then 1500 variables. Trying to organize the ideas on how and what to explore we came to a point where we were drawing causal diagrams such as: A -------&gt; B ------&gt; Outcome | ^ ^ | | | D -------&gt; C ---------- Im not able to draw the diagram we came to, but it looks like it has at least 3 stages and some variables may come into more then one stage simultaneously. And also it seems that at least one variable is outside the diagram and directly related to the outcome. Looking at this diagram it does not seem reasonable to me the a multinomial or a logitic regression may represent such an idea. So my guess is that y = x1B1 + .... + xnBn may not reasonably represent the diagram above. At this point I thought that a neural network could work fine as one theoretical argument often seen at the neural networks text is that it may reasonably approximate any function. So the question is: how to statistically test a diagram such as the above? Also, looking forward for any comment and perhaps suggestions for further reading on this topic. Kind regards, Dr. Pedro Emmanuel A. A. do Brasil Curriculum Lattes: http://lattes.cnpq.br/6597654894290806 Instituto de Pesquisa Clínica Evandro Chagas Fundação Oswaldo Cruz Rio de Janeiro - Brasil Av. Brasil 4365, CEP 21040-360, Tel 55 21 3865-9648 email: pedro.brasil-am19qAsr2QmNrImTOV1V8w&lt; at &gt;public.gmane.org email: emmanuel.brasil-Re5JQEeQqe8AvxtiuMwx3w&lt; at &gt;public.gmane.org ---Apoio aos softwares livres www.zotero.org - gerenciamento de referências bibliográficas. www.broffice.org ou www.libreoffice.org - textos, planilhas ou apresentações. www.epidata.dk - entrada de dados. www.r-project.org - análise de dados. www.ubuntu.com - sistema operacional [[alternative HTML version deleted]] </pre> Pedro Emmanuel Alvarenga Americano do Brasil 2012-09-06T13:17:00 http://comments.gmane.org/gmane.comp.lang.r.epi/133 <pre>R friends, One the things that was limiting me to definitely move from 3.1 to new EpiData with new epx and epz format is that I was not able to import epx/epz data form within R. Currently, from a local R users list I was awere of this user made function. So now Im will try to let 3.1 go and gradually transform my files to XML based format. If there is any EpiData user in this list, could any help me test if it works nicely? https://github.com/daudi/Epidata-XML-to-R kind regards, Dr. Pedro Emmanuel A. A. do Brasil Curriculum Lattes: http://lattes.cnpq.br/6597654894290806 Instituto de Pesquisa Clínica Evandro Chagas Fundação Oswaldo Cruz Rio de Janeiro - Brasil Av. Brasil 4365, CEP 21040-360, Tel 55 21 3865-9648 email: pedro.brasil-am19qAsr2QmNrImTOV1V8w&lt; at &gt;public.gmane.org email: emmanuel.brasil-Re5JQEeQqe8AvxtiuMwx3w&lt; at &gt;public.gmane.org ---Apoio aos softwares livres www.zotero.org - gerenciamento de referências bibliográficas. www.broffice.org ou www.libreoffice.org - textos, planilhas ou apresentações. www.epidata.dk - entrada de dados. www.r-project.org - análise de dados. www.ubuntu.com - sistema operacional [[alternative HTML version deleted]] </pre> Pedro Emmanuel Alvarenga Americano do Brasil 2012-07-30T17:04:04 http://comments.gmane.org/gmane.comp.lang.r.epi/132 <pre>Colleagues, Question: How do YOU create multi-page publication ready tables that can still be shared and potentially edited by coworkers who don't use LaTeX? I'm specifically looking for a solution that accomodates grouped column headers and grouped row specifications and can be imported, WITH styling (e.g., lower border below each set of grouped rows), into MS Word or Excel. Background: I am working on a report with other analysts. They use SAS. I use R. The tables for the report get incorporated into a giant Microsoft Word Document. Their tables are relatively simple without row or column groupings. One of my tables for this report is 762 rows and approximately a dozen columns and needs to have column and row groupings. I drank the kool-aid regarding reproducible research and don't want to post-process the table in MS Word or Excel either. Am I missing something obvious? Things I've tried: I am familiar with the xtable, tables, R2HTML, hwriter, Hmisc::latex and Hmisc:html functions/packages. - xtable doesn't support row or column grouping as far as I can tell. - Hmisc::latex, using longtable generates a nice table with grouped columns and headers (except that the cgroup is not included as part of the header on each subsequent page, unless the tex file is hand edited --- I have e-mailed the package mantainer). If my colleagues could work with the pdf or .tex file, this would be addequate. Unfortunately, conversion is required. I can't get Hmisc::html to work directly with the table. Hevea, used by Hmisc::html, will convert the .tex file outputed by Hmisc::latex after hand-editing the .tex file (e.g.: adding \documentclass{article}, \begin{document}, \providecommand{\tabularnewline}{\\}) but the resulting html file loses the rgroup border formatting when imported into MS Word and LibreOffice. LibreOffice also doesn't seem to know how to handle the A0;A0 encoding which indents the cell contents after the rgroup. - OdfWeave::odfTable might work but I'm having issues (e.g. extremely long processing times and inability to resize the rgroup column). I have e-mailed the package mantainer directly. - Ideally, I would like to generate HTML code for an HTML table and css. This would allow more than one cgroup or rgrouping. I mocked up a proof-of-concept table and it looks good in modern browser but the formatting isn't reflected in recent versions of MS Excel, MS Word or LibreOffice when imported. While my question is really about workflow, I have included sessionInfo, an analagous data set and sample code for odfWeave'ing a LibreOffice file. I am not posting the html file so as not to aggravate the spam filters. ----------------- sessionInfo() ----------------- R version 2.15.1 (2012-06-22) Platform: x86_64-apple-darwin11.4.0 (64-bit) locale: [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 attached base packages: [1] stats graphics grDevices utils datasets methods base other attached packages: [1] odfWeave_0.8.1 XML_3.9-4 lattice_0.20-6 loaded via a namespace (and not attached): [1] grid_2.15.1 tools_2.15.1 ----------------- LibreOffice Info: ----------------- LibreOffice 3.5.5.3 Build ID: 7122e39-92ed229-498d286-15e43b4-d70da21 ----------------- test.odt ----------------- &lt;&lt;table, results=xml, echo=FALSE&gt;&gt;= require(odfWeave) odfSetPageStyle(“RlandscapePage”) rGen &lt;- function(n, length, frame = c(LETTERS, letters, 0:9)){ i = 0 x = NULL while (i &lt; n) { x &lt;- c(x,paste(sample(frame, length, replace = TRUE), collapse = "")) i &lt;- i + 1 } return (x) } # create index and provider name df &lt;- data.frame(NHSN = rGen(254, 6, 0:9), providerName = rGen(254, 30, LETTERS)) # for each index/provider name pair there are three rows for Employees, STVs, and LIPs # I'm sure there is a more elegant way of doing this but I couldn't get expand.grid to work for me in this circumstance df1 &lt;- cbind(df, group = rep("Employees", 254)) df2 &lt;- cbind(df, group = rep("STVs", 254)) df3 &lt;- cbind(df, group = rep("LIPs", 254)) # make one long table with each nhsn/provider listed 3 times, once for each personnel group df4 &lt;- rbind(df1, df2, df3) # create random data for body of table df.body &lt;- data.frame(d = sample(c(0:100,NA), 762, replace = TRUE), vaccinated_value = sample(c(0:100,NA), 762, replace = TRUE), vaccinated_pct = sample(c(0:100,NA), 762, replace = TRUE), contraindicated_value = sample(c(0:100,NA), 762, replace = TRUE), contraindicated_pct = sample(c(0:100,NA), 762, replace = TRUE), declined_value = sample(c(0:100,NA), 762, replace = TRUE), declined_pct = sample(c(0:100,NA), 762, replace = TRUE), unknown_value = sample(c(0:100,NA), 762, replace = TRUE), unknown_pct = sample(c(0:100,NA), 762, replace = TRUE)) # add random data from body to row headers df5 &lt;- cbind(df4, df.body) # order rows by nhsn/provider df5 &lt;- df5[order(df5$NHSN),] df5[is.na(df5)] &lt;- "." df5 &lt;- head(df5, 60) df5$nameNHSN &lt;- paste(df5$providerName, ' (', df5$NHSN, ')', sep = '') colnames &lt;- c("Group","Total",rep(c("n","pct"),4)) cgroup = data.frame(colGroupNames = c("","","Vaccinated","Contraindicated","Declined","Unknown"), colSpecs = c(1,1,2,2,2,2)) rgroup = data.frame(rowGroupNames = unique(df5$nameNHSN), rowSpecs = rep(3, length(df5$nameNHSN)/3)) df5Styles &lt;- tableStyles(df5[,3:12], header = colnames, rgroup = rgroup, cgroup = cgroup, useRowNames = FALSE) df5Styles$cgroupCell &lt;- c("noBorder","noBorder","lowerBorder","lowerBorder","lowerBorder","lowerBorder") df5Styles$cell[seq(3,length(df5Styles$cell[,1]),3),] &lt;- rep("lowerBorder", length(df5Styles$cell[1,])) df5Styles$rgroupCell &lt;- rep("lowerBorder", length(df5Styles$rgroupCell)) odfTable(df5[,3:12], styles = df5Styles, colnames = colnames, useRowNames = FALSE, rgroup = rgroup, cgroup=cgroup) &lt; at &gt; ----------------- ----------------- Creating /var/folders/yc/bd663fhj6_bclc7mp0mqlvl80000gn/T//Rtmpn9zzx6/odfWeave24125717262 Copying ~/test.odt Setting wd to /private/var/folders/yc/bd663fhj6_bclc7mp0mqlvl80000gn/T/Rtmpn9zzx6/odfWeave24125717262 Unzipping ODF file using unzip -o test.odt Archive: test.odt extracting: mimetype extracting: meta.xml inflating: settings.xml inflating: content.xml extracting: Thumbnails/thumbnail.png inflating: manifest.rdf creating: Configurations2/popupmenu/ creating: Configurations2/images/Bitmaps/ creating: Configurations2/toolpanel/ creating: Configurations2/statusbar/ creating: Configurations2/toolbar/ creating: Configurations2/progressbar/ creating: Configurations2/menubar/ creating: Configurations2/floater/ inflating: Configurations2/accelerator/current.xml inflating: styles.xml inflating: META-INF/manifest.xml Removing test.odt Creating a Pictures directory Pre-processing the contents Sweaving content.Rnw Writing to file content_1.xml Processing code chunks ... 1 : term xml(label=table) 'content_1.xml' has been Sweaved Removing content.xml Post-processing the contents --- and here it sits for 10 or more minutes --- Thanks, Jeff </pre> Jeffrey Miller 2012-07-24T19:32:49 http://comments.gmane.org/gmane.comp.lang.r.epi/130 <pre>R friends, with some help of a friend and the original author I turn available this function which is helping me in some analysis. Perhaps it may be useful to others... :-) 1. Gwet KL. Computing inter-rater reliability and its variance in the presence of high agreement. Br J Math Stat Psychol. 2008 ;61(Pt 1):29–48. # AC1 statistic for 2 raters special case # table = k x k table which represents table(rater1,rater2), must have equal number of rows and columns # N = population size which will be stick in standard error correction, N=Inf is no correction. # conflev = Confidence Level associated with the confidence interval (0.95 is the default value) AC1 &lt;- function(table,conflev=0.95,N=Inf,print=TRUE){ if(dim(table)[1] != dim(table)[2]){ stop('The table should have the same number of rows and columns!') } n &lt;- sum(table) f &lt;- n/N pa &lt;- sum(diag(table))/n # formula 18 q &lt;- ncol(table) # number of categories pkk &lt;- diag(table)/n pak &lt;- sapply(1:q,function(i)sum(table[i,]))/n pbk &lt;- sapply(1:q,function(i)sum(table[,i]))/n pik &lt;- (pak + pbk)/2 pegama &lt;- (sum(pik*(1-pik)))/(q-1) gama &lt;- (pa - pegama)/(1 - pegama) # AC1 statistics # 2 raters special case variance pkl &lt;- table/n soma &lt;- 0; for(k in 1:q){ for(l in 1:q){ soma &lt;- soma + (pkl[k,l]*((1-(pik[k]+pik[l])/2)^2)) } } vgama &lt;- ((1-f)/(n*(1-pegama)^2)) * (pa*(1-pa) - 4*(1-gama)*((1/(q-1))*sum(pkk*(1-pik)) - pa*pegama) + 4*((1-gama)^2) * ((1/((q-1)^2))*soma - pegama^2)) epgama &lt;- sqrt(vgama)# AC1 standard error lcb &lt;- max(0,gama - epgama*qnorm(1-(1-conflev)/2,0,1)) # lower confidence bound ucb &lt;- min(1,gama + epgama*qnorm(1-(1-conflev)/2,0,1)) # upper confidence bound if(print==TRUE){ cat('Raw agreement:',pa,'Chance-independent agreement:',pegama,'\n') cat('Agreement coeficient (AC1):',gama,'AC1 standard error:',epgama,'\n') cat(conflev*100,'% Confidence Interval (AC1): (',lcb,',',ucb,')\n') } invisible(c(pa,pegama,gama,epgama,lcb,ucb)) } # table3 &lt;- matrix(c(118,2,5,0),nrow=2,ncol=2) # AC1(table3) # x &lt;- AC1(table3,print=F) # print(x) Regards, Dr. Pedro Emmanuel A. A. do Brasil Curriculum Lattes: http://lattes.cnpq.br/6597654894290806 Instituto de Pesquisa Clínica Evandro Chagas Fundação Oswaldo Cruz Rio de Janeiro - Brasil Av. Brasil 4365, CEP 21040-360, Tel 55 21 3865-9648 email: pedro.brasil-am19qAsr2QmNrImTOV1V8w&lt; at &gt;public.gmane.org email: emmanuel.brasil-Re5JQEeQqe8AvxtiuMwx3w&lt; at &gt;public.gmane.org ---Apoio aos softwares livres www.zotero.org - gerenciamento de referências bibliográficas. www.broffice.org ou www.libreoffice.org - textos, planilhas ou apresentações. www.epidata.dk - entrada de dados. www.r-project.org - análise de dados. www.ubuntu.com - sistema operacional [[alternative HTML version deleted]] </pre> Pedro Emmanuel Alvarenga Americano do Brasil 2012-05-03T17:58:17 http://comments.gmane.org/gmane.comp.lang.r.epi/127 <pre> Hello! I have a doubt in relation with the use of the cosinor model with R when I have not only multiple components, but also linear, quadratic and cubic trends. I explain the details: In simple cosinor: Y = M + A * cos(2 * π* t / T + φ) + error T knowed: y = M + β* X1 + ϒ* X2 + error,   Where X1 = cos (2 * π* t / T) and X2 = sin(2 * π* t / T)   In this case, I use lm(y ~ X1 + X2) .   For the multiple components case:   y = M + ∑Aj * cos(2 * π * t /Tj + φj) + error   In this case I use:  lm(y ~ X1tot + X2tot)   However, in the generalized model case (linear, quadratic and cubic trends), I don’t know how to analyze with R.   I include the situacion formula:   y = M + α1 * t + α2 * t2 + α3 * t3+ ∑Aj * cos(2 *π* t /Tj + φ2) + error   What function in R can I use, as I used lm for the simple cosinor model and the multiple components model?.   How can I obtain global information and separated information?. I refer to obtain information for each contribution (linear tren, quadratic trend, … and so on).   Thank you very much for your help. [[alternative HTML version deleted]] </pre> Iva P 2012-03-18T10:21:11 http://comments.gmane.org/gmane.comp.lang.r.epi/123 <pre>Dear List, I am attempting to carry out linear excess relative risk modelling for the purposes of a nested case-control study of nuclear industry workers. The majority of similar analyses in the literature have been carried out using EPICURE, SAS or STATA, or have been implemented directly in FORTRAN. For a number of reasons, it would be preferable for me to carry out the analysis in R, but I am having difficulty finding an appropriate R package. Many thanks in advance for any advice on this matter, James Grellier James Grellier PhD CREAL-Centre for Research in Environmental Epidemiology Parc de Recerca Biomèdica de Barcelona (despatx 183.01.A) Doctor Aiguader, 88 | 08003 Barcelona Tel. +34 932 147 345 jgrellier-MstSlu9blcReoWH0uzbU5w&lt; at &gt;public.gmane.org &lt;http://www.creal.cat/&gt; www.creal.cat [[alternative HTML version deleted]] </pre> Grellier, James 2012-01-11T13:08:55 http://comments.gmane.org/gmane.comp.lang.r.epi/122 <pre>Course on Age-Period-Cohort models. =================================== Place: Lisbon Time: Monday 19th - Wednesday 21st September 2011. Deadline: 1 September 2011 Fee: 380 EUR. Teacher: Bendix Carstensen, Steno Diabetes Center, Copenhagen URL: http://www.ceaul.fc.ul.pt/seeevent.html?id=206 http://staff.pubhealth.ku.dk/~bxc/APC/Lisboa-2011 ( The latter contains links to various material on the topic ) Audience: Epidemiologists and Statisticians. Age-Period-Cohort models for mortality and morbidity rates is a useful tool for description of age- and time-trends. But the use and in particular the reporting is not straightforward. The course will cover both the basics of tabulation of data, modeling and choice of parametrization as well as the more advanced aspects of Age-Period-Cohort models, including a substantial practical component, based on computer exercises in R. Specifically the apc-tools in the Epi package for R will be covered in some depth. Formally no prerequisites are required, but familiarity with regression models in epidemiology and some exposure to R will prove advantageous. Participants are expected to bring their own laptop, and can expect to go home with a couple of thorough analyses of example datasets. Instructions on what to install (and how) will be given shortly after the application deadline. _________________________________________ Bendix Carstensen Senior Statistician Steno Diabetes Center A/S Niels Steensens Vej 2-4 DK-2820 Gentofte Denmark +45 44 43 87 38 (direct) +45 30 75 87 38 (mobile) bxc-77mEMlKuB9s&lt; at &gt;public.gmane.org www.biostat.ku.dk/~bxc www.steno.dk </pre> BXC (Bendix Carstensen 2011-07-23T12:07:31 http://comments.gmane.org/gmane.comp.lang.r.epi/121 <pre>Dear epidemiological data analysis and R experts, cloudnumbers.com provides researchers and companies with the resources to perform high performance calculations in the cloud. As cloudnumbers.com's community manager I may invite you to register and test R on a computer cluster in the cloud for free: http://my.cloudnumbers.com/register Our aim is to change the way of research collaboration is done today by bringing together scientists and businesses from all over the world on a single platform. cloudnumbers.com is a Berlin (Germany) based international high-tech startup striving for enabling everyone to benefit from the High Performance Computing related advantages of the cloud. We provide easy access to applications running on any kind of computer hardware: from single core high memory machines up to 1000 cores computer clusters. Our platform provides several advantages: * Turn fixed into variable costs and pay only for the capacity you need. Watch our latest saving costs with cloudnumbers.com video: http://www.youtube.com/watch?v=ln_BSVigUhg&amp;feature=player_embedded * Enter the cloud using an intuitive and user friendly platform. Watch our latest cloudnumbers.com in a nutshell video: http://www.youtube.com/watch?v=0ZNEpR_ElV0&amp;feature=player_embedded * Be released from ongoing technological obsolescence and continuous maintenance costs (e.g. linking to libraries or system dependencies) * Accelerated your R, C, C++, Fortran, Python, ... calculations through parallel processing and great computing capacity - more than 1000 cores are available and GPUs are coming soon. * Share your results worldwide (coming soon). * Get high speed access to public databases. * We have developed a security architecture that meets high requirements of data security and privacy. Read our security white paper: http://d1372nki7bx5yg.cloudfront.net/wp-content/uploads/2011/06/cloudnumberscom-security.whitepaper.pdf This is only a selection of our top features. To get more information check out our web-page (http://www.cloudnumbers.com/) or follow our blog about cloud computing, HPC and HPC applications (with R): http://cloudnumbers.com/blog Register and test for free now at cloudnumbers.com: http://my.cloudnumbers.com/register We are looking forward to get your feedback and consumer insights. Best Markus </pre> Markus Schmidberger 2011-07-07T18:11:25 Search the mailing list at Gmane query http://search.gmane.org/?group=$group=gmane.comp.lang.r.epi