Re: [R] Understanding classes in R
On Sun, Sep 29, 2013 at 10:48 PM, john doe anon.r.u...@gmail.com wrote: I am having trouble understanding how classes in R work. Here is a small reproducable example: x=1 class(x) [1] numeric OK. When a variable is a number, its class is numeric. Does R have multiple types for numbers, like C++ (eg integer, float, double). If so, where can I see a list, and how does numeric fit into this system? x=1:100 class(x) [1] integer Wait - I thought that I assigned x to be an array/vector of 100 integers (numerics). Why is the class not array or vector. How is integer different than numeric? Is there a vector or array class in R? If so, why is this not that? In most programming languages scalars and vectors (aka 1-d arrays) are completely different things. However in R a scalar is the same thing as a length-1 vector. Don't think of x=1 and x=1:100 as the first creating a scalar and the second creating a vector containing 100 scalar values. The first creates a vector containing 1 scalar value, and the second creates a vector conatining 100 scalar values. You can't really get scalar values, they'll always effectively be in a vector of length 1. So the return value of class here is actually short for 'vector of numeric' or 'vector of integer' - even when the length is 1 - and you can think of those as the basic numeric classes. There is no 'scalar numeric' class, all is vectors. is.vector(1) is TRUE. There isn't even an is.scalar function. None of that is true for 2-d arrays, aka matrices, where class(m) is always matrix whether its a matrix of characters or numbers. You have to look at the mode(m) or typeof(m) (or storage.mode(m)) to figure out what kind of thing a matrix 'm' contains. A 'list' is a bit more like some of the generic container classes that you find in other programming langages. Its elements can be anything but its class is always 'list'. Use it when you want a vector (in the general sense of 'vector') of non-scalar values, eg L = list(c(1,2,3),1,c(99,120),c(foo,bar,baz)) Confused? Well, just forget everything you learned about classes in your Comp Sci lessons and get ready to learn R's two incompatible OO-programming systems (S3 and S4 classes), or four if you want to look at even more OO systems people have implemented as add-on packages Barry __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] barplot to the right side
Hi, does anybody know how to get a barplot with the x axis starting in the right side and the y axis in the right side too? An example: dat=c(2,4,0,6,5) names(dat)=paste('dpt.',1:5,sep='') barplot(dat, horiz=T) box() I need this barplot in mirror, with the zero value to 6 starting in the right side of x axes and the horizontal bars starting in the right side too. Thank's in advance Juan A. Hernandez-Cabrera [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] climstats
I have been trying to download the climstats package: https://r-forge.r-project.org/R/?group_id=861 but it doesn't seem to run on R 3.0.2 or 3.0.1 and the zipfile is empty. Does anyone know the status of this package or where I can download it. Thanks ** Jenny Williams Spatial Information Scientist, GIS Unit Herbarium, Library, Art Archives Directorate Royal Botanic Gardens, Kew Richmond, TW9 3AB, UK Tel: +44 (0)208 332 5277 email: jenny.willi...@kew.orgmailto:jenny.willi...@kew.org ** Film: The Forgotten Home of Coffee - Beyond the Gardenshttp://www.youtube.com/watch?v=-uDtytKMKpAsns=tw Stories: Coffee Expedition - Ethiopiahttp://storify.com/KewGIS/coffee-expedition-ethiopia Blog: Discovering Coffee in Ethiopia http://www.kew.org/news/kew-blogs/incrEdibles-food-blog/discovering-coffee.htm Kew in Harapan Rainforest Sumatrahttp://storify.com/KewGIS/kew-in-harapan-rainforest Articles: Seeing the wood for the treeshttp://www.kew.org/ucm/groups/public/documents/document/kppcont_060602.pdf How Kew's GIS team and South East Asia botanists are working to help conserve and restore a rainforest in Sumatra. Download a pdf of this article here.http://www.kew.org/ucm/groups/public/documents/document/kppcont_060602.pdf The Royal Botanic Gardens, Kew is a non-departmental public body with exempt charitable status, whose principal place of business is at Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, United Kingdom. The information contained in this email and any attachments is intended solely for the addressee(s) and may contain confidential or legally privileged information. If you have received this message in error, please return it immediately and permanently delete it. Do not use, copy or disclose the information contained in this email or in any attachment. Any views expressed in this email do not necessarily reflect the opinions of RBG Kew. Any files attached to this email have been inspected with virus detection software by RBG Kew before transmission, however you should carry out your own virus checks before opening any attachments. RBG Kew accepts no liability for any loss or damage which may be caused by software viruses. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] barplot to the right side
On 09/30/2013 07:47 PM, Juan Andres Hernandez wrote: Hi, does anybody know how to get a barplot with the x axis starting in the right side and the y axis in the right side too? An example: dat=c(2,4,0,6,5) names(dat)=paste('dpt.',1:5,sep='') barplot(dat, horiz=T) box() I need this barplot in mirror, with the zero value to 6 starting in the right side of x axes and the horizontal bars starting in the right side too. Hi Juan, Try this: par(mar=c(5,2,4,4)) barpos-barplot(-dat,names.arg=rep(,5),axes=FALSE,horiz=TRUE) axis(1,at=(-6):0,labels=6:0) mtext(names(dat),side=4,at=barpos,line=1) Jim __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Interperting results of glmnet and coxph plot, Brier score and Harrel's C-Index - am I doing something wrong ???
To elaborate on Frank's response, the analysis plan of 1. Look at the data and select important variables 2. Put that truncated list into your favorite statistic procedure 3. Ask - are the p-values (c-statistic, coefficients, .) reliable? is a very old plan. The answer to the last question is always NO. The traditional step 1 involved graphs and t-tests, but replacing it by something modern does not change the overall outcome. Terry T __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Apply function to do pairwise calculation
Hello, I want to do pairwise calculation, but I am not sure how to do so. i.e. I have a correlation matrix M 200*200. Namely colnames(M)=rownames(M). In addition, colnames(M) is one of A, B, C, D. I want to first sort the matrix M into 16 modules according to colnames and rownames, and then apply: avecor - function(x,y) { z - (sum(cor(x,y)*cor(x,y))/length(cor(x,y)))^0.5 return(z) } So as to calculate the average correlation between A,B; A,C; A,D; B,C; B,D; C,D. Thanks in advance for your help! Best, Amanda [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Residuals in garch (tseries)
Hello, i'm currently working with the 'garch' function provided by the 'tseries' package in R. If you want to fit a time series you can call the function this way fit = garch(data, order=c(1,1)). A GARCH model delivers you a vector of sd's sigma and therefore confidence intervals for your data. You can get them from the function output by calling fit$fitted.values. You can also get some kind of residuals with fit$residuals and here is my question: does anybody know how they are computed? Normally some would say residuals = data - model, but for GARCH modelling there is no exact model for the data but the confidence interval mentioned above. To compare the residuals I got here from the GARCH modelling to other models I really need to know how they are computed, but the R documentary doesn't provide an answer. Anyone got a clue? Thanks for any help in advance. -- View this message in context: http://r.789695.n4.nabble.com/Residuals-in-garch-tseries-tp4677241.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] change specific factor level values to NA in data frame
Dear R-users I am trying to replace specific factor level values in a data frame with NAs. The data frame includes different kind of variables (e.g, characters, numbers, and factors). I'd like to replace all 'Not applicable', 'Invalid', 'and Missing' for NA. For example: f.level - c('Yes', 'No', 'Not applicable', 'Invalid', 'Missing') df - data.frame(x1=runif(100), x2=sample(f.level, 100, replace=T), x3=sample(f.level, 100, replace=T)) I try changing the values by df[df %in% c('Not applicable', 'Invalid', 'Missing'), ] - NA but nothing seems to change summary(df) My data frame has many more factors. Any advice? Thank you, Daniel __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] barplot - easy for experienced, difficult for me
hey guys, I wanna make a simple barplot, looking like this excel graph: http://r.789695.n4.nabble.com/file/n4677251/barplot_invertebrates.jpg my data set includes 9 groups of invertebrates (x-axes) and total number (y-axes) from two different parks (1 and 2). my data-set looks like that: http://r.789695.n4.nabble.com/file/n4677251/barplot_invertebrates.jpg I'm a bloody beginner and happy for your help. ps: if you know what other graph could be interesting to make out of this very simple data, go ahead and let me know! -- View this message in context: http://r.789695.n4.nabble.com/barplot-easy-for-experienced-difficult-for-me-tp4677251.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] get mouse position without waiting for a click
Consider the following: par(mar=c(0,0,0,0),xaxs = 'i',yaxs='i') plot.new() for(i in 1:20) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } I would like to continuously display the animation until a mouse click happens. Obviously, locator() can't be used for this, because it will halt the animation while waiting for the mouse click. So something like this won't work: while((r-locator(n=1)) == NULL) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } Can somebody suggest a way to get user input while displaying the animation? I am willing to use a keypress if not the mouse. Essentially, I need an input routine that does not wait for a response but just checks at one instant (like kbhit in C). Thanks very much for any help. Stan [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Split type in the RandomForest package
Hi guys, I'm new to Random Forest package and I'd like to know what type of split is used in the package for classification? Or can I configure the package to use different split type (like simple split alongside single attribute axis or linear split based on several attributes etc..) Thanks a lot! [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] how to compare two different models with different dependent variables
i want to compare two exchange rate models namely balassa samuelson model n monetary model of exchange rate.but i don't know how??because both models are different n dependent variables are also different from each other...plz help me in my research work n tell me which econometric technique is usefull in this context? [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] get mouse position without waiting for a click
On Windows or X11, (others ???) ?getGraphicsEvent Cheers, Bert On Mon, Sep 30, 2013 at 7:46 AM, Stanislav Aggerwal stan.agger...@gmail.com wrote: Consider the following: par(mar=c(0,0,0,0),xaxs = 'i',yaxs='i') plot.new() for(i in 1:20) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } I would like to continuously display the animation until a mouse click happens. Obviously, locator() can't be used for this, because it will halt the animation while waiting for the mouse click. So something like this won't work: while((r-locator(n=1)) == NULL) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } Can somebody suggest a way to get user input while displaying the animation? I am willing to use a keypress if not the mouse. Essentially, I need an input routine that does not wait for a response but just checks at one instant (like kbhit in C). Thanks very much for any help. Stan [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] how to compare two different models with different dependent variables
1. This is an r-help list, not a statistics or econometrics list. So you need to post elsewhere. 2. However, having said that, you might find that the CRAN Econometrics task view may be useful. http://cran.r-project.org/web/views/Econometrics.html Cheers, Bert On Mon, Sep 30, 2013 at 4:13 AM, saira khalid saira8...@gmail.com wrote: i want to compare two exchange rate models namely balassa samuelson model n monetary model of exchange rate.but i don't know how??because both models are different n dependent variables are also different from each other...plz help me in my research work n tell me which econometric technique is usefull in this context? [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] DEoptim inconsistent output
On 29 September 2013 19:17, Aya Anas aa...@feps.edu.eg wrote: doesn't make sense at all. I got parameters that don't satisfy the constraint. In addition, when i substituted with the resulting parameters Maybe you have issues with your interpretation of the usage of the package and expected output. So, Can you provide reproducible code? Verbal descriptions may not be efficient to explain issues with mathematical software... __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] change specific factor level values to NA in data frame
Hello, A possibility is the following. icol - sapply(df, is.factor) df[icol] - lapply(df[icol], function(x){ x[as.character(x) %in% c('Not applicable', 'Invalid', 'Missing')] - NA x}) Hope this helps, Rui Barradas Em 30-09-2013 10:42, Daniel Caro escreveu: Dear R-users I am trying to replace specific factor level values in a data frame with NAs. The data frame includes different kind of variables (e.g, characters, numbers, and factors). I'd like to replace all 'Not applicable', 'Invalid', 'and Missing' for NA. For example: f.level - c('Yes', 'No', 'Not applicable', 'Invalid', 'Missing') df - data.frame(x1=runif(100), x2=sample(f.level, 100, replace=T), x3=sample(f.level, 100, replace=T)) I try changing the values by df[df %in% c('Not applicable', 'Invalid', 'Missing'), ] - NA but nothing seems to change summary(df) My data frame has many more factors. Any advice? Thank you, Daniel __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Residuals in garch (tseries)
Hello, To know how something in R is computed, download the sources and read the respective function code. That's one of the benefits of open source software, you can know exactly what is going on. Hope this helps, Rui Barradas Em 30-09-2013 10:31, ZuckerRahmen escreveu: Hello, i'm currently working with the 'garch' function provided by the 'tseries' package in R. If you want to fit a time series you can call the function this way fit = garch(data, order=c(1,1)). A GARCH model delivers you a vector of sd's sigma and therefore confidence intervals for your data. You can get them from the function output by calling fit$fitted.values. You can also get some kind of residuals with fit$residuals and here is my question: does anybody know how they are computed? Normally some would say residuals = data - model, but for GARCH modelling there is no exact model for the data but the confidence interval mentioned above. To compare the residuals I got here from the GARCH modelling to other models I really need to know how they are computed, but the R documentary doesn't provide an answer. Anyone got a clue? Thanks for any help in advance. -- View this message in context: http://r.789695.n4.nabble.com/Residuals-in-garch-tseries-tp4677241.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] change specific factor level values to NA in data frame
Well, maybe or maybe not. The problem is that the old factor levels remain. Here's a tiny example that illustrates the issue: z - factor(c(a,b,c)) z[as.character(z)==c] - NA z [1] abNA Levels: a b c Whether or how you wish to change this depends on what you are doing with the data. Cheers, Bert On Mon, Sep 30, 2013 at 9:52 AM, Rui Barradas ruipbarra...@sapo.pt wrote: Hello, A possibility is the following. icol - sapply(df, is.factor) df[icol] - lapply(df[icol], function(x){ x[as.character(x) %in% c('Not applicable', 'Invalid', 'Missing')] - NA x}) Hope this helps, Rui Barradas Em 30-09-2013 10:42, Daniel Caro escreveu: Dear R-users I am trying to replace specific factor level values in a data frame with NAs. The data frame includes different kind of variables (e.g, characters, numbers, and factors). I'd like to replace all 'Not applicable', 'Invalid', 'and Missing' for NA. For example: f.level - c('Yes', 'No', 'Not applicable', 'Invalid', 'Missing') df - data.frame(x1=runif(100), x2=sample(f.level, 100, replace=T), x3=sample(f.level, 100, replace=T)) I try changing the values by df[df %in% c('Not applicable', 'Invalid', 'Missing'), ] - NA but nothing seems to change summary(df) My data frame has many more factors. Any advice? Thank you, Daniel __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] climstats
On Sep 30, 2013, at 3:25 AM, Jenny Williams wrote: I have been trying to download the climstats package: https://r-forge.r-project.org/R/?group_id=861 but it doesn't seem to run on R 3.0.2 or 3.0.1 What makes you say this? What errors are reprorted? (Doesn't seems to run is a bit vague.) and the zipfile is empty. I was able to install version 1.0 from sources with: install.packages(climstats, repos=http://R-Forge.R-project.org;, type=source) (I agree that the zipfile for Windows was not found.) R version 3.0.1 Patched (2013-07-23 r63392) Running Mac OS 10.7.5. It appears to require a fair number of external package, so you would need to check the Depends in the description file. Depends: R (= 2.13), raster, rgdal, chron, zoo, sp, ncdf, R.utils It did not appear to do any C or Fortran compiling, so I think that means you do not need to have RTools installed on Windows. But since it requires rgdal, you would need to have GDAL installed if you were to get it to load. Does anyone know the status of this package or where I can download it. Thanks ** Jenny Williams Spatial Information Scientist, GIS Unit Herbarium, Library, Art Archives Directorate Royal Botanic Gardens, Kew Richmond, TW9 3AB, UK Tel: +44 (0)208 332 5277 email: jenny.willi...@kew.orgmailto:jenny.willi...@kew.org ** Film: The Forgotten Home of Coffee - Beyond the Gardenshttp://www.youtube.com/watch?v=-uDtytKMKpAsns=tw Stories: Coffee Expedition - Ethiopiahttp://storify.com/KewGIS/coffee-expedition-ethiopia Blog: Discovering Coffee in Ethiopia http://www.kew.org/news/kew-blogs/incrEdibles-food-blog/discovering-coffee.htm Kew in Harapan Rainforest Sumatrahttp://storify.com/KewGIS/kew-in-harapan-rainforest Articles: Seeing the wood for the treeshttp://www.kew.org/ucm/groups/public/documents/document/kppcont_060602.pdf How Kew's GIS team and South East Asia botanists are working to help conserve and restore a rainforest in Sumatra. Download a pdf of this article here.http://www.kew.org/ucm/groups/public/documents/document/kppcont_060602.pdf The Royal Botanic Gardens, Kew is a non-departmental public body with exempt charitable status, whose principal place of business is at Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, United Kingdom. The information contained in this email and any attachments is intended solely for the addressee(s) and may contain confidential or legally privileged information. If you have received this message in error, please return it immediately and permanently delete it. Do not use, copy or disclose the information contained in this email or in any attachment. Any views expressed in this email do not necessarily reflect the opinions of RBG Kew. Any files attached to this email have been inspected with virus detection software by RBG Kew before transmission, however you should carry out your own virus checks before opening any attachments. RBG Kew accepts no liability for any loss or damage which may be caused by software viruses. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Interperting results of glmnet and coxph plot, Brier score and Harrel's C-Index - am I doing something wrong ???
On Sep 29, 2013, at 2:16 PM, E Joffe wrote: HI, Thank you for your answer. There were 301 events out of 394 observations. Study goals: Identify proteins with prognostic power in patients with AML. There were 232 proteins studied. Traditional models won't converge. I wanted to do a multivariate survival analysis that would allow me to identify prognostic proteins and get an estimation of their hazard ratio. LASSO allowed me to model the data but in order to get the HRs I needed to run it a second time throw coxph - this was done based on a discussion on stackexchange http://r.789695.n4.nabble.com/estimating-survival-times-with-glmnet-and-coxph-td4614225.html which I tweaked a bit. The features selected by LASSO are not necessarily all significant when constructing a coxph model in particular due to the lack of penalization. This was commented on by Dr. Hastie in one of his papers (or in the video lecture he gave - I need to find the exact source). I understand that by doing this procedure I might be reducing the accuracy of the model which is why I used a bootstrap evaluation. Interestingly, the coxph models following stepwise feature selection had marginally better performance in the evaluation compared with the LASSO models (maybe the LASSO over-fit the data a bit ? - don't know) What I still don't understand is why with a C-Index of 0.76 and a Brier score of 0.07 when I look at the plot of the coxph object the confidence intervals are so far ? The item you referenced was not on stackexchange but rather on Nabble. The archived copy in R-help is here: [R] Estimating survival times with glmnet and coxph https://stat.ethz.ch/pipermail/r-help/2012-May/312029.html https://stat.ethz.ch/pipermail/r-help/2012-May/312038.html The response from Terry Therneau at that time explained why the CI's were wide. I think this implies that even without the additional (faulty) step of stepwise reduction in covariates you still would not be able to use the CI's coxphfrom such a fit. -- David. Further, I am wondering whether there is a sanity check I could perform to make sure there is no error in the Brier/C-Index evaluation. Thanks again for all you advice on this project, Erel -Original Message- From: David Winsemius [mailto:dwinsem...@comcast.net] Sent: Saturday, September 28, 2013 4:41 PM To: E Joffe Cc: r-help@r-project.org Subject: Re: [R] Interperting results of glmnet and coxph plot, Brier score and Harrel's C-Index - am I doing something wrong ??? On Sep 28, 2013, at 2:39 AM, E Joffe wrote: Hi all, I am using COX LASSO (glmnet / coxnet) regression to analyze a dataset of 394 obs. / 268 vars. I use the following procedure: 1. Construct a coxnet on the entire dataset (by cv.glmnet) 2. Pick the significant features by selecting the non-zero coefficient under the best lambda selected by the model 3. Build a coxph model with bi-directional stepwise feature selection limited to the coxnet selected features. I was a bit puzzled by the third step. Once you had a reduced model from glmnet, what was the statistical basis for further elimination of variables? (Quite apart from the statistical issues, I was rather surprised that this procedure even produced results since the 'step' function is not described in the 'stats' package as applying to 'coxph' model objects.) To validate the model I use both Brier score (library=peperr) and Harrel's [Harrell] C-Index (library=Hmisc) with a bootstrap of 50 iterations. MY QUESTION : I am getting fairly good C-Index (0.76) and Brier (0.07) values for the models however per the coxnet the %Dev explained by the model is at best 0.27 and when I plot the survfit of the coxph the plotted confidence interval is very large. How many events did you have? (The width of CI's is most importantly dependent on event counts and not particularly improved by a high case count. The power considerations are very similar to those of a binomial test.) What am I missing here ? Perhaps sufficient events? (You also seem to be missing a description of the study goals.) -- David. %DEV=27% Brier score - 0.07 ($ipec.coxglmnet - [1] 7.24) C-Index - 0.76 ($cIndex - [1] 0.763) DATA: [Private Health Information - can't publish] 394 obs./268 vars. CODE (need to define a dataset with 'time' and 'status' variables): library(survival) library (glmnet) library (c060) library (peperr) library (Hmisc) #creat Y (survival matrix) for glmnet surv_obj - Surv(dataset$time,dataset$status) ## tranform categorical variables into binary variables with dummy for dataset predict_matrix - model.matrix(~ ., data=dataset, contrasts.arg = lapply (dataset[,sapply(dataset, is.factor)], contrasts)) ## remove the statu/time variables from the predictor matrix (x) for glmnet
[R] FW: Library update from version
Hello, I recently installed version 3.0.1 of R on to a computer. I have a working installation for a Statconn application using R version 2.15.0 on another computer. I have many libraries under this old installation. Can I just copy them into the new library from the old, or do I install each one of them under the new R? Also how can I get a list of differences in two libraries so that I can use this list to update the new one? Thank you. Cem __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] get mouse position without waiting for a click
Some code that you can look at for examples of capturing the mouse position without clicking include the playSudoku function in the sudoku package and several functions in the TeachingDemos package including HWidentify, HTKidentify, dynIdentify, and TkIdentify. On Mon, Sep 30, 2013 at 8:46 AM, Stanislav Aggerwal stan.agger...@gmail.com wrote: Consider the following: par(mar=c(0,0,0,0),xaxs = 'i',yaxs='i') plot.new() for(i in 1:20) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } I would like to continuously display the animation until a mouse click happens. Obviously, locator() can't be used for this, because it will halt the animation while waiting for the mouse click. So something like this won't work: while((r-locator(n=1)) == NULL) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } Can somebody suggest a way to get user input while displaying the animation? I am willing to use a keypress if not the mouse. Essentially, I need an input routine that does not wait for a response but just checks at one instant (like kbhit in C). Thanks very much for any help. Stan [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Gregory (Greg) L. Snow Ph.D. 538...@gmail.com [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] FW: Library update from version
Hi, On Mon, Sep 30, 2013 at 11:12 AM, Cem Girit gi...@comcast.net wrote: Hello, I recently installed version 3.0.1 of R on to a computer. I have a working installation for a Statconn application using R version 2.15.0 on another computer. I have many libraries under this old installation. Can I just copy them into the new library from the old, or do I install each one of them under the new R? No, you shouldn't do that. Also how can I get a list of differences in two libraries so that I can use this list to update the new one? A bit of googling could have provided several answers. This post in particular has a few answers from some people who know what they're doing w/ R, so probably a good place to start: http://stackoverflow.com/questions/1401904/painless-way-to-install-a-new-version-of-r-on-windows HTH, -steve -- Steve Lianoglou Computational Biologist Bioinformatics and Computational Biology Genentech __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] ks.test or other Kolmogorov-Smirnov code
Can any of the tools available provide where the supremum happens? So classically D = sup|F1 - F2| or something to that affect, and then we use D to figure out significance. If the ecdf is to be read as y=Pr(xx_i), for each x_i , well at which x_i does D occur? Sorry for the repetition just trying to give all the context of what I'm asking. Thanks in advance. -- View this message in context: http://r.789695.n4.nabble.com/ks-test-or-other-Kolmogorov-Smirnov-code-tp4677290.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] get mouse position without waiting for a click
Very good question -- sorry to have left that info out of my posting. windows() Thanks for any help. Bill On Mon, Sep 30, 2013 at 4:46 PM, Bert Gunter gunter.ber...@gene.com wrote: On Windows or X11, (others ???) ?getGraphicsEvent Cheers, Bert On Mon, Sep 30, 2013 at 7:46 AM, Stanislav Aggerwal stan.agger...@gmail.com wrote: Consider the following: par(mar=c(0,0,0,0),xaxs = 'i',yaxs='i') plot.new() for(i in 1:20) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } I would like to continuously display the animation until a mouse click happens. Obviously, locator() can't be used for this, because it will halt the animation while waiting for the mouse click. So something like this won't work: while((r-locator(n=1)) == NULL) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } Can somebody suggest a way to get user input while displaying the animation? I am willing to use a keypress if not the mouse. Essentially, I need an input routine that does not wait for a response but just checks at one instant (like kbhit in C). Thanks very much for any help. Stan [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ks.test or other Kolmogorov-Smirnov code
If the ecdf is to be read as y=Pr(xx_i), for each x_i , well at which x_i does D occur? The definition is Pr(x = x_i), but that doesn't matter much here. You know the maximum occurs at one of the points in union(x, y) so you can find it by comparing the ecdf's at each of those points: f - function(x, y) { xy - union(x, y) d - abs(ecdf(x)(xy) - ecdf(y)(xy)) xy[ d == max(d) ] } Bill Dunlap Spotfire, TIBCO Software wdunlap tibco.com -Original Message- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of sitterlyb Sent: Monday, September 30, 2013 10:39 AM To: r-help@r-project.org Subject: [R] ks.test or other Kolmogorov-Smirnov code Can any of the tools available provide where the supremum happens? So classically D = sup|F1 - F2| or something to that affect, and then we use D to figure out significance. If the ecdf is to be read as y=Pr(xx_i), for each x_i , well at which x_i does D occur? Sorry for the repetition just trying to give all the context of what I'm asking. Thanks in advance. -- View this message in context: http://r.789695.n4.nabble.com/ks-test-or-other- Kolmogorov-Smirnov-code-tp4677290.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] get mouse position without waiting for a click
thanks Greg Bill On Monday, September 30, 2013, Greg Snow 538...@gmail.com wrote: Some code that you can look at for examples of capturing the mouse position without clicking include the playSudoku function in the sudoku package and several functions in the TeachingDemos package including HWidentify, HTKidentify, dynIdentify, and TkIdentify. On Mon, Sep 30, 2013 at 8:46 AM, Stanislav Aggerwal stan.agger...@gmail.com javascript:_e({}, 'cvml', 'stan.agger...@gmail.com'); wrote: Consider the following: par(mar=c(0,0,0,0),xaxs = 'i',yaxs='i') plot.new() for(i in 1:20) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } I would like to continuously display the animation until a mouse click happens. Obviously, locator() can't be used for this, because it will halt the animation while waiting for the mouse click. So something like this won't work: while((r-locator(n=1)) == NULL) { z - matrix(runif(256*256), ncol=256) dev.hold() image(z, col=grey(0:255/255),zlim=c(0,1),useRaster=TRUE) dev.flush() Sys.sleep(.1) } Can somebody suggest a way to get user input while displaying the animation? I am willing to use a keypress if not the mouse. Essentially, I need an input routine that does not wait for a response but just checks at one instant (like kbhit in C). Thanks very much for any help. Stan [[alternative HTML version deleted]] __ R-help@r-project.org javascript:_e({}, 'cvml', 'R-help@r-project.org');mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. -- Gregory (Greg) L. Snow Ph.D. 538...@gmail.com javascript:_e({}, 'cvml', '538...@gmail.com'); [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] FW: Library update from version
In this case the OP probably does need to reinstall contributed packages since going from 2.15.x to 3.0.y entails a major version change in R. Dennis On Mon, Sep 30, 2013 at 11:49 AM, Steve Lianoglou lianoglou.st...@gene.com wrote: Hi, On Mon, Sep 30, 2013 at 11:12 AM, Cem Girit gi...@comcast.net wrote: Hello, I recently installed version 3.0.1 of R on to a computer. I have a working installation for a Statconn application using R version 2.15.0 on another computer. I have many libraries under this old installation. Can I just copy them into the new library from the old, or do I install each one of them under the new R? No, you shouldn't do that. Also how can I get a list of differences in two libraries so that I can use this list to update the new one? A bit of googling could have provided several answers. This post in particular has a few answers from some people who know what they're doing w/ R, so probably a good place to start: http://stackoverflow.com/questions/1401904/painless-way-to-install-a-new-version-of-r-on-windows HTH, -steve -- Steve Lianoglou Computational Biologist Bioinformatics and Computational Biology Genentech __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] FW: Library update from version
On Mon, Sep 30, 2013 at 2:49 PM, Steve Lianoglou lianoglou.st...@gene.com wrote: Hi, On Mon, Sep 30, 2013 at 11:12 AM, Cem Girit gi...@comcast.net wrote: Hello, I recently installed version 3.0.1 of R on to a computer. I have a working installation for a Statconn application using R version 2.15.0 on another computer. I have many libraries under this old installation. Can I just copy them into the new library from the old, or do I install each one of them under the new R? No, you shouldn't do that. Really? Why not? Note that the Windows upgrade FAQ (http://cran.r-project.org/bin/windows/base/rw-FAQ.html#What_0027s-the-best-way-to-upgrade_003f) says to do exactly that. Best, Ista Also how can I get a list of differences in two libraries so that I can use this list to update the new one? A bit of googling could have provided several answers. This post in particular has a few answers from some people who know what they're doing w/ R, so probably a good place to start: http://stackoverflow.com/questions/1401904/painless-way-to-install-a-new-version-of-r-on-windows HTH, -steve -- Steve Lianoglou Computational Biologist Bioinformatics and Computational Biology Genentech __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] ks.test or other Kolmogorov-Smirnov code
sweet, I didn't know ecdf could be use like that! Thanks! -- View this message in context: http://r.789695.n4.nabble.com/ks-test-or-other-Kolmogorov-Smirnov-code-tp4677290p4677306.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] barplot - easy for experienced, difficult for me
Look at the example for grouped bar plots here, http://www.statmethods.net/graphs/bar.html Jean On Mon, Sep 30, 2013 at 8:05 AM, happyR tobi_gebetsber...@gmx.at wrote: hey guys, I wanna make a simple barplot, looking like this excel graph: http://r.789695.n4.nabble.com/file/n4677251/barplot_invertebrates.jpg my data set includes 9 groups of invertebrates (x-axes) and total number (y-axes) from two different parks (1 and 2). my data-set looks like that: http://r.789695.n4.nabble.com/file/n4677251/barplot_invertebrates.jpg I'm a bloody beginner and happy for your help. ps: if you know what other graph could be interesting to make out of this very simple data, go ahead and let me know! -- View this message in context: http://r.789695.n4.nabble.com/barplot-easy-for-experienced-difficult-for-me-tp4677251.html Sent from the R help mailing list archive at Nabble.com. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Apply function to do pairwise calculation
Amanda, If I understand what you're trying to do, this example might help you. M - structure(c(0.66, 0.05, -0.93, -0.61, 0.65, -0.25, 0.23, -0.89, 0.37, 0.38, -0.91, 0.91, -0.05, -0.65, -0.94, 0.73, -0.88, 0.25, 0.04, -0.89, -0.47, -0.46, 0.86, -0.29, 0.92, 0.22, 0.77, -0.98, -0.56, 0.11, 0.35, -0.49, 0.48, 0.42, -0.28, 0.16, 0.2, 0.61, 0.7, 0.94, 0.89, 0.46, 0.93, 0.36, -0.95, -0.01, 0.27, 0.73, -0.17, -0.93, 0.36, 0.78, 0.02, -0.8, -0.57, -0.03, -0.81, -0.74, 0.16, 0.52, 0.65, -0.45, 0.46, -0.68, -0.85, 0.74, 0.77, -0.39, 0.03, 0.93, -0.23, -0.91, -0.14, 0.77, -0.36, 0.43, -0.5, 0.5, -0.68, -0.6, 0.12, -0.18, 0.16, -0.38, -0.18, -0.59, -0.3, -0.14, 0.53, -0.53, -0.56, 0.14, -0.54, 0.93, 0.54, 0.03, 0.06, 0.14, 0.97, -0.09), .Dim = c(10L, 10L), .Dimnames = list(c(D, C, D, C, B, A, C, A, C, B), c(D, C, D, C, B, A, C, A, C, B))) meanM - data.frame(row=NA, col=NA, mean=NA) count - 0 for(i in 1:(ng-1)) { for(j in (i+1):ng) { count - count + 1 meanM[count, row] - groups[i] meanM[count, col] - groups[j] meanM[count, mean] - mean(M[dimnames(M)[[1]]==groups[i], dimnames(M)[[2]]==groups[j]]) }} Jean On Mon, Sep 30, 2013 at 9:55 AM, Amanda Li amand...@uchicago.edu wrote: Hello, I want to do pairwise calculation, but I am not sure how to do so. i.e. I have a correlation matrix M 200*200. Namely colnames(M)=rownames(M). In addition, colnames(M) is one of A, B, C, D. I want to first sort the matrix M into 16 modules according to colnames and rownames, and then apply: avecor - function(x,y) { z - (sum(cor(x,y)*cor(x,y))/length(cor(x,y)))^0.5 return(z) } So as to calculate the average correlation between A,B; A,C; A,D; B,C; B,D; C,D. Thanks in advance for your help! Best, Amanda [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] multilevel analysis
I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] FW: Library update from version
Hi, On Mon, Sep 30, 2013 at 12:44 PM, Ista Zahn istaz...@gmail.com wrote: On Mon, Sep 30, 2013 at 2:49 PM, Steve Lianoglou lianoglou.st...@gene.com wrote: Hi, On Mon, Sep 30, 2013 at 11:12 AM, Cem Girit gi...@comcast.net wrote: Hello, I recently installed version 3.0.1 of R on to a computer. I have a working installation for a Statconn application using R version 2.15.0 on another computer. I have many libraries under this old installation. Can I just copy them into the new library from the old, or do I install each one of them under the new R? No, you shouldn't do that. Really? Why not? Because it has been my experience that people *just* do that (ie. the *just* copy the libraries, as the OP asked). But as you correctly point out: Note that the Windows upgrade FAQ (http://cran.r-project.org/bin/windows/base/rw-FAQ.html#What_0027s-the-best-way-to-upgrade_003f) says to do exactly that. There is a way to copy the old package and then ensure that they are updated to the versions that build correctly on the newest version of R. -steve -- Steve Lianoglou Computational Biologist Bioinformatics and Computational Biology Genentech __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] multilevel analysis
On Sep 30, 2013, at 2:50 PM, srecko joksimovic wrote: I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? But normal distribution is not a requirement for linear models. Please review your theory. thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] multilevel analysis
I thought so, but then I found this: Normality The assumption of normality states that the error terms at every level of the model are normally distributed maybe I misinterpreted something. On Mon, Sep 30, 2013 at 3:06 PM, David Winsemius dwinsem...@comcast.netwrote: On Sep 30, 2013, at 2:50 PM, srecko joksimovic wrote: I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? But normal distribution is not a requirement for linear models. Please review your theory. thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] multilevel analysis
-Original Message- From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On Behalf Of srecko joksimovic Sent: Monday, September 30, 2013 3:22 PM To: David Winsemius Cc: R help Subject: Re: [R] multilevel analysis I thought so, but then I found this: Normality The assumption of normality states that the error terms at every level of the model are normally distributed maybe I misinterpreted something. Yes, error terms (i.e. residuals), not your raw data, should be normally distributed if you are going to conduct significance tests that rely on the assumption of normality. But as David said, the assumption of normality (in any form) is not necessary for estimating linear models. Dan On Mon, Sep 30, 2013 at 3:06 PM, David Winsemius dwinsem...@comcast.netwrote: On Sep 30, 2013, at 2:50 PM, srecko joksimovic wrote: I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? But normal distribution is not a requirement for linear models. Please review your theory. thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA Daniel Nordlund Bothell, WA USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] multilevel analysis
On Sep 30, 2013, at 3:22 PM, srecko joksimovic wrote: I thought so, but then I found this: Normality The assumption of normality states that the error terms at every level of the model are normally distributed maybe I misinterpreted something. Notice that it is the _error_terms_ that are to be normally distributed, not the data itself. One might even infer that normally distrited data might be suspect because the correct distribution should be a mixture of normals. Since the errors never are going to fit on a straight line on a QQ plot, the real question is how far from Normal and what the impact might be on the quantities being estimated. -- David. On Mon, Sep 30, 2013 at 3:06 PM, David Winsemius dwinsem...@comcast.net wrote: On Sep 30, 2013, at 2:50 PM, srecko joksimovic wrote: I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? But normal distribution is not a requirement for linear models. Please review your theory. thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA David Winsemius Alameda, CA, USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] multilevel analysis
Thanks for your comments, David and Bert. The best would be to provide an example. Let's say we have a dataset like this one: IDEmployee Company OU CountViewPortal CountLogin TimeOnTask Performance 1 Company1 Company1.OU1 21 33 627.8 4.3 2 Company1 Company1.OU2 45 54 34.8 2.3 3 Company2 Company1.OU1 23 33 3.8 1.0 4 Company2 Company1.OU1 34 12 44.8 2.3 5 Company2 Company1.OU2 55 22 55.8 4.5 6 Company2 Company1.OU3 45 44 34.8 3 I want to see if there is correlation between CountViewPortal and Performance. Moreover, I'd like to reveal the influence of CountViewPortal+TimeOnTask on Performance. However, I expect that employees within a OU, and than a Company have similar behavior. Thus, I'll have 3 levels - employee, OU, Company. In R, I would do something like this: randomInterceptCount - lme(Performance ~ CountViewPortal, data=analysis, random=~1|OU/Company1, method=ML) But, then the point is that CountViewPortal, CountLogin and TimeOnTask are non-normally distributed. I guess that my question is, what should I do in case of non-normal distribution? I really appreciate your help. Thanks again! Srecko On Mon, Sep 30, 2013 at 5:14 PM, David Winsemius dwinsem...@comcast.netwrote: On Sep 30, 2013, at 3:22 PM, srecko joksimovic wrote: I thought so, but then I found this: Normality The assumption of normality states that the error terms at every level of the model are normally distributed maybe I misinterpreted something. Notice that it is the _error_terms_ that are to be normally distributed, not the data itself. One might even infer that normally distrited data might be suspect because the correct distribution should be a mixture of normals. Since the errors never are going to fit on a straight line on a QQ plot, the real question is how far from Normal and what the impact might be on the quantities being estimated. -- David. On Mon, Sep 30, 2013 at 3:06 PM, David Winsemius dwinsem...@comcast.net wrote: On Sep 30, 2013, at 2:50 PM, srecko joksimovic wrote: I have an example of multilevel analysis with 3 levels, but data are non-normally distributed. In case of normal distribution, I would perform multilevel linear analysis using lme function, but what should I do in case of non-normal distribution? But normal distribution is not a requirement for linear models. Please review your theory. thanks, Srecko [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. David Winsemius Alameda, CA, USA David Winsemius Alameda, CA, USA [[alternative HTML version deleted]] __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] [sm.density.compare] scale up y-axis and additional line type
Dear fellows, The two questions are on sm.density.compare(). I compare kernel density estimates of two arrays of data. I'd like to scale up y-axis so that I can show better the differences in y values. English is not my first language so I'll try to explain it. I would like to stretch y-axis a bit longer but not to change the range of y values. How can I do this? Second, I'm using a solid line for one array of date and a dashed line for another array of data. The difference is not easy to see, because the two curves are approximate to some degree. Is it possible to plot one array of data with *, instead of dash? Thanks! Kind regards, Xianwen __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.