[R] nls
Hello, I want to estimate parameters of the model y=a/x with x=1:10 and y=3/x. I tested the NLS function. Could you tell me why it doesn't converge ? Regards, Grégory Benmenzer x=1:10 y=3/x nls(y~a/x, start=list(a=2), control=nls.control(maxiter=100)) Error in nls(y ~ a/x, start = list(a = 2), control = nls.control(maxiter = 100)) : number of iterations exceeded maximum of 100 __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] DSC 2003: Workshop on Distributed Statistical Computing
Dear R users, the preliminary program for the 3rd international workshop on 'Distributed Statistical Computing' (DSC 2003) is now available at http://www.ci.tuwien.ac.at/Conferences/DSC-2003/ The workshop will take place at the Technische Universität Wien in Vienna, Austria from March 20-22. The deadline for registrations is March 14, but the deadline for early registration ends on February 14. Further details can be found on the webpage. This workshop will deal with future directions in (open source) statistical computing and graphics. Topics of particular interest include: * Bioinformatics * Database Connectivity * Graphical Modeling * User Interfaces and Office Integration * Resample and Combine Methods * Spatial Statistics * Visualization Emphasis will be given to the R, Omegahat and BioConductor projects. In addition to the conference program there will be several half-day tutorials on March 19 about: * An Introduction to BioConductor * Exploring Genomic Data Using R and BioConductor * Writing R Extensions * R Graphics For the organization committee Achim Zeileis Achim Zeileis Institut für Statistik Technische Universität Wien http://www.ci.tuwien.ac.at/~zeileis/ __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] shapiro.test
Hi The shapiro.test function outputs a value of the W statistic, which should be 1 if the distribution is normal, and a p-value for the test (as the documentation states). I'm a bit confused with some results. I'm getting a W=0.9977 and a p-value=0.1889. I was expecting that a W of 0.9977 would tell me that the distribution is normal so p-value should be small ... What am I missing ? Thanks EJ __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] shapiro.test
Ernesto Jardim wrote: Hi The shapiro.test function outputs a value of the W statistic, which should be 1 if the distribution is normal, and a p-value for the test (as the documentation states). I'm a bit confused with some results. I'm getting a W=0.9977 and a p-value=0.1889. I was expecting that a W of 0.9977 would tell me that the distribution is normal so p-value should be small ... What am I missing ? Thanks EJ __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help You have it backwards. The null hypothesis is that the distribution is Normal. You reject this null when the p-value is small. If the distribution is Normal, the p-value will tend to be large. shapiro.test(rnorm(100)) Shapiro-Wilk normality test data: rnorm(100) W = 0.9877, p-value = 0.4894 Rick B. __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] problems using lqs()
On 10 Feb 2003 at 15:40, Luca Scrucca wrote: I am not sure about this, but lqs have an extra argument `adjust'. from the help page: adjust: should the intercept be optimized for each sample? Defaults to TRUE When you use formula, the estimation algorithm knows the columns of 1's is the intercept, and can traet it accordingly. When yo make the matrix X with on regressor column being all1's, the estimation algorithm does'nt know it represents the intercept (you did'nt give it a name). Could this be the reason why the different results? Testing this conjecture: mod4 - lqs(y ~ x1 + x2, method=lms, nsamp=exact, adjust=FALSE) coef(mod4) (Intercept) x1 x2 35.4217275 0.4276641 -1.2834731 coef(mod1) (Intercept) x1 x2 35.5293489 0.4422742 -1.2944534 coef(mod2) x1 x2 35.4217275 0.4276641 -1.2834731 so this might be the explanation. By the way, for the last question, using debug() and looking at the output gives: Browse[1] n debug: subset - eval(substitute(subset), data, env) Browse[1] n debug: data - .Internal(model.frame(formula, rownames, variables, varnames, extras, extranames, subset, na.action)) Browse[1] n Error in model.frame(formula, rownames, variables, varnames, extras, extranames, : variable lengths differ so the error occurs in model.frame. You could continue the debugging yourself. (don't have time now). Kjetil Halvorsen Dear List-members, I found a strange behaviour in the lqs function. Suppose I have the following data: y - c(7.6, 7.7, 4.3, 5.9, 5.0, 6.5, 8.3, 8.2, 13.2, 12.6, 10.4, 10.8, 13.1, 12.3, 10.4, 10.5, 7.7, 9.5, 12.0, 12.6, 13.6, 14.1, 13.5, 11.5, 12.0, 13.0, 14.1, 15.1) x1 - c(8.2, 7.6,, 4.6, 4.3, 5.9, 5.0, 6.5, 8.3, 10.1, 13.2, 12.6, 10.4, 10.8, 13.1, 13.3, 10.4, 10.5, 7.7, 10.0, 12.0, 12.1, 13.6, 15.0, 13.5, 11.5, 12.0, 13.0, 14.1) x2 - c(23.005, 23.873, 26.417, 24.868, 29.895, 24.200, 23.215, 21.862, 22.274, 23.830, 25.144, 22.430, 21.785, 22.380, 23.927, 33.443, 24.859, 22.686, 21.789, 22.041, 21.033, 21.005, 25.865, 26.290, 22.932, 21.313, 20.769, 21.393) and I would like to fit the following model: mod1 - lqs(y ~ x1 + x2, method=lms, nsamp=exact) mod1$coefficients (Intercept) x1 x2 35.5293489 0.4422742 -1.2944534 mod1$bestone [1] 12 17 27 Now, instead of using the formula, I want to provide the design matrix and the response, then: X - cbind(1, x1, x2) mod2 - lqs.default(X, y, intercept=F, method=lms, nsamp=exact) mod2$coefficients x1 x2 35.4217275 0.4276641 -1.2834731 mod2$bestone [1] 6 14 15 The results are not the same (?!). Furthermore, if I create the design matrix without the column of 1's for the intercept: X - cbind(x1, x2) mod3 - lqs.default(X, y, intercept=T, method=lms, nsamp=exact) mod3$coefficients (Intercept) x1 x2 35.5293489 0.4422742 -1.2944534 mod3$bestone [1] 12 17 27 I get the same result I had using the formula (see mod1). This is confusing me! Another small problem appears if I re-fit the first model using the formula and asking to return the x-matrix and the response: lqs(y ~ x1 + x2, method=lms, nsamp=exact, x.ret=T, y.ret=T) Error in model.frame(formula, rownames, variables, varnames, extras, extranames, : variable lengths differ Does anyone know what's going on? Regards, with thanks in advance, Luca Scrucca +---+ | Dr. Luca Scrucca | | Dipartimento di Scienze Statistiche tel. +39 - 075 - 5855278 | | Universita' degli Studi di Perugia fax. +39 - 075 - 43242 | | Via Pascoli - C.P. 1315 Succ. 1 | | 06100 PERUGIA (ITALY)| | | | E-mail: [EMAIL PROTECTED] | | Web page: http://www.stat.unipg.it/luca | +---+ __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] Zero rows/cols in the hessian matrix
Hi, | From: Timur Elzhov [EMAIL PROTECTED] | Date: Mon, 10 Feb 2003 19:06:18 +0300 | | Dear R experts! | | I try to minimize a function with external C fitting function. | I get the hessian matrix. Here it is: | | [,1] [,2] [,3] [,4] | [1,] 1.88166310 0.88598030 | [2,] 0.0000 0.0000 | [3,] 0.88598030 0.48599830 | [4,] 0.0000 0.0000 First, are you sure that your fitting (minimisation?) routine handles the problem correctly? Not all of the optimising routines are able to deal with constant parameters. | Second and fourth rows/columns have zero values only. That's OK, | because that ones related to parameters were not included in fitting | expression (but *were* passed to minimization function as arguments), | so dF/dp == 0 for them. It of course doesn't make sense to calculate | the standard errors for the mentioned fitting-independent | parameters, but I want to do that for others! I read in R-intro, that | I have to calculate the inverse of hessian at first. | solve(hessian) logs: | | Error in solve.default(hessian) : singular matrix `a' in solve The matrix is definitely singular. In the example above, you have in fact fitted a 2-parameter model and made a hessian which includes 2 extra rows and columns (of course, it depends on your fitting algorithm, but I guess it handles the problem in this way). Then you have to exclude those rows and columns when you invert the hessian. I use to do so: ind - c(1, 3) # indices you need varcov - matrix(0, 4, 4) varcovar[ind,ind] - solve(hessian[ind,ind]) I.e. invert only parameter-depending part of the hessian and put it into the corresponding elements in the full varcovar matrix (if you need that). Be sure how your fitting algorithm handles constant parameters! Best wishes, Ott | So, the question is: | How can I calculate the errors of remaining parameters (without | removing fitting-independent parameters from arguments)? | | Thanks! | | | -- | WBR, | Timur __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] lda
R per se does not have lda() function. Package MASS does, and MASS (the book) describes it in detail. If you use a package supporting a book (there are several) do expect to read the book for the fine details On Mon, 10 Feb 2003, Luis Silva wrote: There are some versions of lda. I would like to know which one is behind R's lda function. -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] problems using lqs()
Many thanks to Kjetil Halvorsen for explaining me the reason I got
different results.
Regarding the second point, Both Kjetil and Brian Ripley suggest to go
through debugging and see what the reason for the message.
I'm not so expert to fully understand why, but I have found a patch that
makes the function works. It is based on the behaviour of the lm function.
The original function is:
lqs.formula - function (formula, data, ..., method = c(lts, lqs,
lms, S, model.frame), subset, na.action = na.fail, model = TRUE,
x.ret = FALSE, y.ret = FALSE, contrasts = NULL)
{
method - match.arg(method)
...
if (x.ret)
fit$x - x
if (y.ret)
fit$y - y
fit
}
I only changed the name of the arguments (x.ret to x, y.ret to y) as in
lm():
lqs.formula - function (formula, data, ..., method = c(lts, lqs,
lms, S, model.frame), subset, na.action = na.fail, model = TRUE, x =
FALSE, y = FALSE, contrasts = NULL)
{
ret.x - x
ret.y - y
method - match.arg(method)
...
if (ret.x)
fit$x - x
if (ret.y)
fit$y - y
fit
}
Note that we need also to replace the 'x' with something else in:
lqs - function(what, ...) UseMethod(lqs)
Now
mod - lqs(y ~ x1 + x2, x=T, y=T)
and mod$x and mod$y returns what expected. The x-matrix is without the
enventual column of 1's. Maybe we may want to add it if required.
I hope this may help to fix the problem.
Best,
Luca
+---+
| Dr. Luca Scrucca |
| Dipartimento di Scienze Statistiche tel. +39 - 075 - 5855278 |
| Universita' degli Studi di Perugia fax. +39 - 075 - 43242 |
| Via Pascoli - C.P. 1315 Succ. 1 |
| 06100 PERUGIA (ITALY)|
| |
| E-mail: [EMAIL PROTECTED] |
| Web page: http://www.stat.unipg.it/luca |
+---+
__
[EMAIL PROTECTED] mailing list
http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] (no subject)
my name is Tomer Maimon, and i'm a student in the israeli Institue for science ( Technion) .i have some problem with the R program. i want to write formula with two factors to predict(in rpart algorithm). i'll be more then thankfull if you could send me an answer internet link, where can i find answers to this kind of problems. __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] non-SQL sqlQuery error
Dear all, I've encountered a curious problem. I am trying to run an SQL query using sqlQuery() function in RODBC. The query works fine when run in a stand-alone SQL browser (Microsoft Query Analyzer, in particular). However, when I use the exact same thing from sqlQuery() function, I get the following error: Error in [.data.frame(data, , ) : not all specified columns exist Traceback() gives this: 5: stop(not all specified columns exist) 4: [.data.frame(data, , ) 3: data[, ] 2: sqlGetResults(channel, errors = errors, ...) 1: sqlQuery(con.object, ... long query here ... ) The puzzling thing is that the error does not seem to be like the usual mispecified-SQL type that I normally get when using sqlQuery(). I am not sure how to debug this problem, since I can't really see what's going on in the sqlQuery() internally. My query includes a SUM(ColumnName1) GROUP BY ColumnName2 construct, and from trial and error, it seems that this is the cause of the failure (at least removing it makes the error go away). I am using R1.6.2 and the RODBC package from http://cran.r-project.org/bin/windows/contrib/PACKAGES. Can anyone offer any clues? Thanks, Pijus __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] non-SQL sqlQuery error
On Mon, 10 Feb 2003, Pijus Virketis wrote: I've encountered a curious problem. I am trying to run an SQL query using sqlQuery() function in RODBC. The query works fine when run in a stand-alone SQL browser (Microsoft Query Analyzer, in particular). However, when I use the exact same thing from sqlQuery() function, I get the following error: Error in [.data.frame(data, , ) : not all specified columns exist Traceback() gives this: 5: stop(not all specified columns exist) 4: [.data.frame(data, , ) 3: data[, ] 2: sqlGetResults(channel, errors = errors, ...) 1: sqlQuery(con.object, ... long query here ... ) There is no statement 3 in the released version of sqlGetResults ... The puzzling thing is that the error does not seem to be like the usual mispecified-SQL type that I normally get when using sqlQuery(). I am not sure how to debug this problem, since I can't really see what's going on in the sqlQuery() internally. You can. R are RODBC and Open Source, so why not open it and read it? -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] Type of multi-valued variable
Hi, I've read in the past a thead in the R discussion list about the multi-valued type variable (what was called checklist). At the moment Gregory had intention to add some general code in his gregmisc package. I'm wondering if there's some general code / packages available ? A general class for taking account this type of variable would be very useful in the domain of survey processings, as multi-responses questions are often used. The simple operations applied to these variables are holecount, cross tabulations with others variables, transformation to single coded variables like number of responses, etc. Thanks in advance for any help -- Fan __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] multilm for simseg/acm
On Mon, 10 Feb 2003, Christian Schulz wrote: hi, for working with the simseg/acm approach i need multilm and it seems that the last version is 0.1-4.tar ? What's wrong in the description file, if i attempt install the package ? c:\DataMining\rw1062\binRcmd Install multilm Malformed DCF file (file multilm/DESCRIPTION, line 6) [R1.6.2 /W2K] P.S. ..hmm a solution might be import the functions from multilm in my .Rprofile, but i would like understand the package problem ! It appears to have an empty Depends: line, so delete it. Also, the wrapping on the Description: line is wrong, and there is no Maintainer field. -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272860 (secr) Oxford OX1 3TG, UKFax: +44 1865 272595 __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] geoR question from new R user
Hi, I'm a new R user. My goal is to do a variogram using geoR. I started by trying to do the example in the geoR Illustrative Session using my own data. I am able to read in my Ascii data using: D - matrix(scan(file.dat, n=530*3), 530,3, byrow=TRUE). Then I use: as.geodata(D, coords.col=1:2, data.col=3) to make the object D geodata. I check the descriptive statistics, and my data file appears to be in order. But when I try the following command: cloud1- variog(D, option=cloud, max.dist=1), I get this error: Error in array(x, c(length(x), 1), if (!is.null(names(x))) list (names(x), : attempt to set an attribute on NULL I get that same error message when I try different arguments with the variog command. Can someone help? Thanks, Sharon __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
Re: [R] installation on FreeBSD
Dr Mccarter and Dr Gentleman, Thank you for your reply. I reinstalled the system and found out that you can not install R on FreeBSD 5.0 (I mean in the default setting when installing R and also I tried to open the --enable-R-shlib. I chose installing full binary, kernel src, X during the installation of FreeBSD) but can install R on FreeBSD 4.7. I don't have ability to figure out why. Just bring it out. Best wishes, Peng *** Peng Zhang Department of Biostatistics Harvard School of Public Health 655 Huntington Avenue Boston, Massachusetts 02115 *** I believe I can fly I believe I can touch the sky On Sun, 9 Feb 2003, loren mccarter wrote: I'm not sure what is causing it to dump core. I'm using FreeBSD 4.7 and I'm waiting for 5.1 before upgrading so I can't troubleshoot it on my end. Here are some things you may want to try: (1)~use /stand/sysinstall to add R as a binary package (R-1.5.1_1 is available this way), (2)~if you think it's X, try configuring R to install without X (I think there is a configure option for this, something like: configure --without-x). (3)~when you installed the FreeBSD packages, make sure that none of them had XFree86 3.3.6 as a dependency, which would cause it to overwrite the 4.2 libraries (but, you'd really notice if this happened this b/c X would not work at all, not just R). (4)~if this is a brand new machine setup, you may want to use the more stable FreeBSD 4.7 until 5.1 comes out in a few months. Of course, none of those suggestions are ideal, but may help you to trouble shoot the issue. Loren On Sun, 9 Feb 2003, Peng Zhang wrote: Dear loren, Thank you for your reply! I did try to install directly from source code. However I got the following error information: ../../../../library/methods/libs/methods.so is unchanged dumping R code in package 'methods' Segmentation fault (core dumped) *** Error code 139 Stop in /usr/home/pzhang/R-1.6.2/src/library/methods. *** Error code 1 Whether is there something wrong with my X settings? Thank you for your help. Best wishes, Peng *** Peng Zhang Department of Biostatistics Harvard School of Public Health 655 Huntington Avenue Boston, Massachusetts 02115 *** I believe I can fly I believe I can touch the sky On Sun, 9 Feb 2003, loren mccarter wrote: I've been using R on FreeBSD for several years now so I can tell you it works fine. However I have not yet upgraded to FreeBSD 5.0 so perhaps it has something to do with the new version. Here's something you may want to try: Rather than using the old version of R from the FreeBSD ports collection, download the latest R source directly from CRAN then try to ./configure make make install it. I've found that R obeys the FreeBSD 'hier' and the sources on CRAN are always more up to date than the compliled FreeBSD packages or ports. Loren On Sat, 8 Feb 2003, Peng Zhang wrote: Hello there, I just changed to FreeBSD platform, and want to install R on it. I use FreeBSD 5.0 and install nearly all packages on the machine. When I use ports to install R, (cd /usr/ports/math/R-letter, and then type make) I got the following error information. ../../../../library/methods/libs/methods.so is unchanged dumping R code in package 'methods' Fatal error: The X11 shared library could not be loaded. The error was /usr/ports/math/R-letter/work/R-1.6.0/modules/R_X11.so: Undefined symbol R_GlobalEnv Can somebody tell me why and how to fix this problem? Thank you for your help! Best wishes, Peng *** Peng Zhang Department of Biostatistics Harvard School of Public Health 655 Huntington Avenue Boston, Massachusetts 02115 *** I believe I can fly I believe I can touch the sky __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
[R] RPgSQL W2K
Hi, i found tis message in the archive and have got the same problems ? John, perhaps you have found now a way install RPgSQL for windows2000, or anybody other ? many thanks for advance christian [SNIP] I wonder whether anyone has succeeded in building either the RPgSQL or the Rdbi.PGSQL and Rdbi packages for Windows. If so, would you be willing to share either the binary package(s) or instructions about how to set up the configure scripts? My preliminary efforts to get things going under Windows 2000 haven't met with much success. Thanks, John [SNIP] [[alternate HTML version deleted]] __ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
