Re: [R] small sample size confidence interval by bootstrap
On Fri, 31 Mar 2006, Urania Sun wrote: I only have 4 samples. I wish to get a confidence interval around the mean. Is it reasonable? If not, is there a way to compute a confidence interval for such small sample size's mean? (BTW, the CI is for the population mean, not the sample mean. I'll also assume that you are prepared to assume that you have a single random sample of size 4 from a location family.) For a confidence interval, you need to make some assumptions about the distribution. If you assume normality, you can use t.test, but the estimate of the standard deviation (on just 3 df) will be very variable and this will be reflected in the length of the CI. Your subject line mentions the bootstrap. You could use one of several different types of bootstrap CI but they also make assumptions, weaker assumptions that lead to even more variability. For a sample of size 4 there are (at most) 36 distinct means of bootstrap resamples, so none of the methods I know of will work adequately (and most not at all). As an example to ponder, the Cauchy distribution does not even have a mean, but from small samples you will have no idea that is very long-tailed. And getting a CI for a location parameter is often better done from a robust estimator of location than from the sample mean. Alternatively, your true distribution might be a discrete distribution on 5 points, and you have no idea at all about the 5th value. -- 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 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] [S] S-PLUS 8 beta program [repost]
On 3/31/06, eugene dalt [EMAIL PROTECTED] wrote: In addition, I think people should always explore ways to repackage exclusive S-Plus facilities and libraries into R. By rewritting them or using GPL loopholes. Licenses are important. They state for the author, precisely what the author intends. Just because it's not really what the author might want, is a different story. The GPL has no loopholes. It does precisely what it is supposed to do, which it to provide free-as-in-speech software, which maintains that freedom. Read it; love it; use it :-). I do. Take BioConductor, which isn't GPL (it would've been by my choice, which is what I argued for when we released it the first time). The intent was for others to use it, and not be forced to share the changes. I went along with that, fully aware of what I was going to be contributing to. It seems to me that Insightful is very good at protecting whatever they create and the same time feels very comfortable taking R stuff to keep they clients happy. In essence they are selling free stuff. Packaging is worth something. It's why folks pay for Linux distributions. I'm very happy that Insightful packages up free stuff. I wish they'd do it better and add incredible amounts of value to S-PLUS, so that it could truly compete with R. There are some really nice facilities that R doesn't have, and probably will never have. But it's still missing key features. And I'm happy that they are finally listening to comments (I can't claim to be the only one that suggested this over 7 years ago -- I also can't claim to be a user (again), until my new job; but it's nice that it took them less than a decade). best, -tony [EMAIL PROTECTED] Muttenz, Switzerland. Commit early,commit often, and commit in a repository from which we can easily roll-back your mistakes (AJR, 4Jan05). __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] [S] S-PLUS 8 beta program [repost]
On 3/31/06, A.J. Rossini [EMAIL PROTECTED] wrote: There are some really nice facilities that R doesn't have, and probably will never have. But it's still missing key features. It's late, and somewhere in my 18+ hour day today, I spent a miserable evening in Heathrow. Yech. I meant: There are some really nice facilities in S-PLUS that R doesn't have, and probably never will have. But S-PLUS is still missing key features. best, -tony [EMAIL PROTECTED] Muttenz, Switzerland. Commit early,commit often, and commit in a repository from which we can easily roll-back your mistakes (AJR, 4Jan05). __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Disable LazyLoading mechanism completely
On Sat, 1 Apr 2006, Prof Brian Ripley wrote: On Fri, 31 Mar 2006, Dirk Eddelbuettel wrote: On 31 March 2006 at 16:15, Paul Roebuck wrote: | Is there a global option somewhere that can completely | disable the LazyLoad option? I want all my packages | in source format for searching purposes, crippled by | the conversion to database format. AFAICT setting LazyLoad=No in DESCRIPTION might do the trick. For reasons I fail to understand I have to set it to that value for Debian builds of the Rcmdr package, and the resulting package does indeed have R code as plain sources. I'd agree with you that being able to read, and hence search, the code is a good thing but I also trust the R masters on the advantages of lazy loading. There may be an option to diable it globally ... Quite a large number of packages (including almost all using S4 methods) *do not work* without lazyloading or saved images. So this option would just break other people's work. Can you elaborate a bit on this statement? S4 predates LazyLoad and used to work just fine before that mechanism was introduced. Indeed, several of my own packages have S4 methods and work with neither LazyLoading nor saved images. Should that be read as since package authors made one-way changes to support LazyLoad? Is there some way to tell whether a package can do without it? -- SIGSIG -- signature too long (core dumped) __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] reference paper about SOM
On eof the best sources of info on the SOM is Kohonen's book. Try and
get this if you can.
@Book{kohonen95:_self,
author = {Teuvo Kohonen},
title ={Self-organizing maps},
publisher ={Springer},
year = 1995,
address = {Berlin}
}
On 01/04/06, Linda Lei [EMAIL PROTECTED] wrote:
Hi All,
I'm looking for some reference paper about SOM (self organizing map)
algorithm. I tried the paper which is mentioned in
the help page of function som (package:som):
http://www.cis.hut.fi/research/papers/som_tr96.ps.Z
But I can't open it for some reason. Could you please help me with it ?
Thanks a lot!
[[alternative HTML version deleted]]
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Disable LazyLoading mechanism completely
On 3/31/2006 6:03 PM, Paul Roebuck wrote: On Fri, 31 Mar 2006, Duncan Murdoch wrote: On 3/31/2006 5:15 PM, Paul Roebuck wrote: Is there a global option somewhere that can completely disable the LazyLoad option? I want all my packages in source format for searching purposes, crippled by the conversion to database format. Why not just keep the source? Maybe I misunderstood where you're thinking of searching, but I think even without LazyLoad, you'll lose comments in the functions. My thought is something along the lines of options(lazy.load = FALSE) which would override the setting in the DESCRIPTION file as though every package had set it as: LazyLoad: no I want the ability to patch/search the code by default. LazyLoad removes this capability. I know there will be a minor speed hit and accept that in exchange for the above capability. Lazy loading just replaces each object with a promise to load it, right? So if you need to do a search/patch, couldn't you force those promises during the search? Duncan Murdoch __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Problem with help() respectively ?
Werner Wernersen wrote: Hi, I have a problem with my the R help system. I have been searching through the help archives but I can't find anything about it and I don't know how to specify my search better. When I type for instance ?hist, just the next command line prompt shows but nothing actually happens. There is no help window coming up. I already tried reinstalling R (2.2.1, windows XP) but that didn't change anything. I am pretty sure the help has been working before on this machine. Has anybody an idea how to recover the help system? Which help format do you use? What about help(help, chmhelp=FALSE, htmlhelp=FALSE, offline=FALSE) Does this work fo you? Uwe Ligges Thanks for your help, Werner __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] rowVars
r user wrote: I am using the R 2.2.1 in a Windows XP environment. I have a dataframe with 12 columns and 1,000 rows. (Some of the rows have 1 or fewer values.) fewer than 1? So you mean 0? Well, the commonly used estimator for the variance is defined for n=2 elements only! I am trying to use rowVars to calculate the variance of each row. I don't know a function rowVars in R-2.2.1 - and cannot find it ... I am getting the following message: “Error in na.remove.default(x) : length of 'dimnames' [1] not equal to array extent” Hard to tell without knowing what you actually did. Uwe Ligges Is there a good work-around? __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] (no answer)
I have never taken a statistics class nor read a statistics text, but
I am in dire need of help with a trivial data analysis problem for
which I need to write a report in two hours. I have spent 10,000
hours of study in my field of expertise (high frequency noise-making
plant biology) but I've always thought that statistics is something
that can be mastered on short notice.
Briefly, I have an experiment in which a response variable is
repeatedly measured at 1-day intervals, except that after a plant
becomes sick, it is measured every three days. We forgot to randomize
on one of the important variables (soil pH) and we forgot to measure
the soil pH. Plants that begin to respond to treatment are harvested
and eaten (deep fried if they don't look so good), but we want to make
an inference about long-term responses. In addition, we forgot to
measure the response on some of the days before the plant was
terminated. Some baseline variables were not measured for some
plants, when some of the other variables looked OK. The response
variable is only known to exceed a certain value in some cases, and in
others is only known to be less than a certain value. The response
variable also has a great number of ties at zero, and has extreme high
outliers. The variability of responses seems to depend on whether
there was missing variables for the plant. And halfway through the
experiment we changed instrumentation and personnel. All of these
problems seem trivial when compared to what I have to deal with every
day in measuring plant sounds, so I hope that someone can help me as
soon as possible. I would appreciate receiving a few paragraphs of
description of the analysis that I can include in my report, and I
would like to receive R code to analyze the data no matter which
variables I collect. I do value your time, so you will get my
everlasting thanks.
Note that I will be out of the office from 1:15pm to 1:25pm today.
This information should be valuable to many.
I. Ben Fuld
Technical University of Plant Kinetics
Slapout, Alabama
LEGAL NOTICE\ Unless expressly stated otherwise, this messag...{{dropped}}
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] (no answer)
On 01-Apr-06 Frank E Harrell Jr wrote:
I have never taken a statistics class nor read a statistics text, but
I am in dire need of help with a trivial data analysis problem for
which I need to write a report in two hours. I have spent 10,000
hours of study in my field of expertise (high frequency noise-making
plant biology) but I've always thought that statistics is something
that can be mastered on short notice.
Dear Ibn Fuld (I apologise for rewriting your name correctly,
and I do appreciate the problems of people who do not natively
speak English, but I thought that doind so would be useful for
the members of the list who have the coverse problem),
You have an evidently complex problem there, but apparently a
very short time in which to solve it. Happily, I see a very
simple solution.
Just talk to your plants, ask them how they are, record their
acoustic responses, and use your existing expertise to analyse
and interpret the latter. I judge that you need to learn nothing
new to do this, your institution should posses the required
technology, and I suspect that little if any new R code would be
required.
Good luck,
ZB [EMAIL PROTECTED]
Briefly, I have an experiment in which a response variable is
repeatedly measured at 1-day intervals, except that after a plant
becomes sick, it is measured every three days. We forgot to randomize
on one of the important variables (soil pH) and we forgot to measure
the soil pH. Plants that begin to respond to treatment are harvested
and eaten (deep fried if they don't look so good), but we want to make
an inference about long-term responses. In addition, we forgot to
measure the response on some of the days before the plant was
terminated. Some baseline variables were not measured for some
plants, when some of the other variables looked OK. The response
variable is only known to exceed a certain value in some cases, and in
others is only known to be less than a certain value. The response
variable also has a great number of ties at zero, and has extreme high
outliers. The variability of responses seems to depend on whether
there was missing variables for the plant. And halfway through the
experiment we changed instrumentation and personnel. All of these
problems seem trivial when compared to what I have to deal with every
day in measuring plant sounds, so I hope that someone can help me as
soon as possible. I would appreciate receiving a few paragraphs of
description of the analysis that I can include in my report, and I
would like to receive R code to analyze the data no matter which
variables I collect. I do value your time, so you will get my
everlasting thanks.
Note that I will be out of the office from 1:15pm to 1:25pm today.
This information should be valuable to many.
I. Ben Fuld
Technical University of Plant Kinetics
Slapout, Alabama
LEGAL NOTICE\ Unless expressly stated otherwise, this
messag...{{dropped}}
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
E-Mail: (Ted Harding) [EMAIL PROTECTED]
Fax-to-email: +44 (0)870 094 0861
Date: 01-Apr-06 Time: 15:44:50
-- XFMail --
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] (no answer)
Oh, and I forgot to add. Please generate some test data for me since
I can't possibly take time out to provide such in order to clarify the
question. By the way, I did try out R a bit but it did not work and its
too much effort to provide the R code I have or to reduce it to a small
self contained reproducible example to illustrate the salient point.
I am thanking you in advance since I will probably be too busy
to acknowledge the help or to summarize the answers for the
benefit of others and the list archives. Its not that I don't want to
but I probably won't follow up on your answers anyways if they involve reading
and thinking about help pages, the manual, the FAQ, the posting guide,
statistics, mathematics, programming or other material.
By the way, please email me directly since I normally don't read
the list.
:)
On 4/1/06, Frank E Harrell Jr [EMAIL PROTECTED] wrote:
I have never taken a statistics class nor read a statistics text, but
I am in dire need of help with a trivial data analysis problem for
which I need to write a report in two hours. I have spent 10,000
hours of study in my field of expertise (high frequency noise-making
plant biology) but I've always thought that statistics is something
that can be mastered on short notice.
Briefly, I have an experiment in which a response variable is
repeatedly measured at 1-day intervals, except that after a plant
becomes sick, it is measured every three days. We forgot to randomize
on one of the important variables (soil pH) and we forgot to measure
the soil pH. Plants that begin to respond to treatment are harvested
and eaten (deep fried if they don't look so good), but we want to make
an inference about long-term responses. In addition, we forgot to
measure the response on some of the days before the plant was
terminated. Some baseline variables were not measured for some
plants, when some of the other variables looked OK. The response
variable is only known to exceed a certain value in some cases, and in
others is only known to be less than a certain value. The response
variable also has a great number of ties at zero, and has extreme high
outliers. The variability of responses seems to depend on whether
there was missing variables for the plant. And halfway through the
experiment we changed instrumentation and personnel. All of these
problems seem trivial when compared to what I have to deal with every
day in measuring plant sounds, so I hope that someone can help me as
soon as possible. I would appreciate receiving a few paragraphs of
description of the analysis that I can include in my report, and I
would like to receive R code to analyze the data no matter which
variables I collect. I do value your time, so you will get my
everlasting thanks.
Note that I will be out of the office from 1:15pm to 1:25pm today.
This information should be valuable to many.
I. Ben Fuld
Technical University of Plant Kinetics
Slapout, Alabama
LEGAL NOTICE\ Unless expressly stated otherwise, this messag...{{dropped}}
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] -newbie | RODBC import query
Greetings - After 20+ years of using SAS, for a variety of reasons, I'm using [R] for a bunch of things - while I'm getting a pretty good a handling [R] for script programming, and statistical analysis, I'm struggling with 'pulling data into [R]'. For reasons beyond my control, a number of the files I get sent to 'work with' are in Dbase format (*.dbf). For another host of reasons, I need to be able to read directly into [R] from these files (no using intermediate .CVS or delimited ASCII files). OK, so after a bit of reading, seems I need to use RODBC (I'm using [R] 2.2.1 for Windows, at the moment). But, I can't seem to figure out the basics. Suppose the file I need to 'work with' is test.dbf So, I try the following: library(RODBC); import_dat - odbcConnectDbase(c:\documents and settings\egc\desktop\test.dbf) OK, so far so good - well, at least no outright errors gets chunked out to the console. Now what? Here's where I get stuck. There is a table in the test.dbf file called TEST. But, the following tester - sqlFetch(import_dat,TEST) blows up - I get the following error message in the console: Error in odbcTableExists(import_dat, sqtable) : 'TEST': table not found on channel OK - so it doesn't seem to find the table TEST in test.dbf. I tried lower-case for TEST (i.e., test), but that doesn't seem to solve the problem. OK, so lets pretend I don't know what the table in test.dbf is called, and use sqlTables instead: table_list - sqlTables(import_dat) When I then enter table_list in the console, I get [1] TABLE_CAT TABLE_SCHEM TABLE_NAME TABLE_TYPE REMARKS 0 rows (or 0-length row.names) Meaning, what? It almost seems that its telling me there is nothing in test.dbf. Well, there definitely is (I can open it up in Excel - shudder), but, perhaps it is unable to recognize whats there. Suggestions? Apologies if this is easy, or (worse) and FAQ. Thanks! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] -newbie | RODBC import query
On 4/1/06, Evan Cooch [EMAIL PROTECTED] wrote: Greetings - After 20+ years of using SAS, for a variety of reasons, I'm using [R] for a bunch of things - while I'm getting a pretty good a handling [R] for script programming, and statistical analysis, I'm struggling with 'pulling data into [R]'. For reasons beyond my control, a number of the files I get sent to 'work with' are in Dbase format (*.dbf). For another host of reasons, I need to be able to read directly into [R] from these files (no using intermediate .CVS or delimited ASCII files). OK, so after a bit of reading, seems I need to use RODBC (I'm using [R] 2.2.1 for Windows, at the moment). But, I can't seem to figure out the basics. Suppose the file I need to 'work with' is test.dbf So, I try the following: library(RODBC); import_dat - odbcConnectDbase(c:\documents and settings\egc\desktop\test.dbf) \ is the escape character so if you want a \ you must use \\ like this: odbcConnectionDbase(c:\\documents and settings\\egc\\desktop\\test.dbf) or do this: odbcConnectionDbase(file.choose()) OK, so far so good - well, at least no outright errors gets chunked out to the console. Now what? Here's where I get stuck. There is a table in the test.dbf file called TEST. But, the following tester - sqlFetch(import_dat,TEST) blows up - I get the following error message in the console: Error in odbcTableExists(import_dat, sqtable) : 'TEST': table not found on channel OK - so it doesn't seem to find the table TEST in test.dbf. I tried lower-case for TEST (i.e., test), but that doesn't seem to solve the problem. OK, so lets pretend I don't know what the table in test.dbf is called, and use sqlTables instead: table_list - sqlTables(import_dat) When I then enter table_list in the console, I get [1] TABLE_CAT TABLE_SCHEM TABLE_NAME TABLE_TYPE REMARKS 0 rows (or 0-length row.names) Meaning, what? It almost seems that its telling me there is nothing in test.dbf. Well, there definitely is (I can open it up in Excel - shudder), but, perhaps it is unable to recognize whats there. Suggestions? Apologies if this is easy, or (worse) and FAQ. Thanks! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] barchart in black white with 10 categories?
Dear list, I am trying to plot a barchart (from lattice) in B W, with 10 categories per bar. It seems that the colours are recycled after reachingcategory number 7, which creates a problem interpreting the chart. I therefore have two questions: 1) How do I get more shades? 2) Does anyone have a theme to share with me with distinctive shades in Black White? Thankful for all the help I can get. /Fredrik __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] -newbie | RODBC import query
At 12:14 PM 4/1/2006, Gabor Grothendieck wrote: OK, so after a bit of reading, seems I need to use RODBC (I'm using [R] 2.2.1 for Windows, at the moment). But, I can't seem to figure out the basics. Suppose the file I need to 'work with' is test.dbf So, I try the following: library(RODBC); import_dat - odbcConnectDbase(c:\documents and settings\egc\desktop\test.dbf) \ is the escape character so if you want a \ you must use \\ like this: odbcConnectionDbase(c:\\documents and settings\\egc\\desktop\\test.dbf) or do this: odbcConnectionDbase(file.choose()) Well, OK, but that doesn't make any apparent difference it in terms of solving the problem I'm having. the odbcConnecDbase command is working fine (i.e., its finding the test.dbf file), but after that, nada. __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] S-PLUS 8 beta program [repost]
This thread may already contain too many contributions, but I feel compelled to add something: STEALING? Insightful was accused of stealing GNU software. On that issue, I will as the accusor(s) and other to please read the GNU license (e.g., at http://www.gnu.org/copyleft/gpl.html), making special not of the following portion of paragraph 2 of TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION: If identifiable sections of that work [software that uses GNU code] are not derived from the Program [GNU code or derivative], and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate work. I work for a company that develops and licenses specialized software for a narrow market niche. We do not currently connect our software to R, but we plan to do so in the future, using a completely separate add-on module that we can distribute free with the source code, consistent with the GNU license. If Insightful is guilty of stealing when making GNU code available to their customers, then I'm guilty of stealing each time I apply something I've learned from a published research report or book. I have not yet contributed any packages to CRAN, though I have contributed suggestions that have improved some of R's capabilities. Moreover, I expect to contribute packages in the future. I would be happy to have Insightful make my code more widely available (especially if I don't have to do any more work to make it available to the wider audience). FOR PROFIT vs. NOT FOR PROFIT Some years ago, I worked for a not-for-profit research firm. I was paid to work there, and my boss carefully explained to me, We're not for profit, but we're not for loss, either. I now work for a company that develops and distributes data analysis software for a narrow, specialized market. Some of our former competitors went out of business, because they did not charge as much as we do and could not afford to develop and maintain their code enough to compete with us. Our customers pay our license fees, because they believe the use of our software (1) saves them substantially more money than we charge and (2) allows them to provide their customers with better products at lower cost. I also like the R model, with (a) many of the R core team being university professors who support R as part of the research obligations of their job and (b) substantive contributions by many others around the world who contribute small portions of their spare time to support this r-help listserve and contribute code to CRAN. I don't agree with the ideologues of either the Right or the Left: For profit, not for profit and volunteer efforts all make positive contributions to building a better world. Best Wishes, spencer graves eugene dalt wrote: Point well taken. You should, however, expect R users to bring up concerns. This isn't a win-win situation as you sound...and you want to keep s-news in the dark too. Frankly, you didn't address the real issue. How would Insightful reacts for example if they find R users repackaging your products named Infact or Insightful Miner? or as you said R users (instead of Splus users) need access to the cutting-edge, quality statistical software available in Splus (instead of other way round). Insightful wants to take any cutting-edge, quality statistical software from R, but they patent any cutting-edge, quality statistical software they create. Hence my call to people to use any loopholes to make these available or rewrite them. That would be win-win. ps. This is my last email on this issue. --- David Smith [EMAIL PROTECTED] wrote: eugene dalt [EMAIL PROTECTED] writes: It seems to me that Insightful is very good at protecting whatever they create and the same time feels very comfortable taking R stuff to keep they clients happy. In essence they are selling free stuff. I must defend Insightful on this point. As an employee of Insightful, that's to be expected, but I've also been an author of free software since the earliest days when the term free software was in use ... but more on that, later. In no sense are we taking R stuff with these improvements in S-PLUS 8. Packages written for R will remain free, as they must. This isn't just because the license says so, but because we believe that an open-source community around packages that will run with both S-PLUS and R is a good thing. It's good for S-PLUS users, certainly: we've heard loud and clear from our users that they need a better way to package user-contributed libraries for S-PLUS, and they need access to the cutting-edge, quality statistical software available in the open-source world. But it's also good for the statistical
Re: [R] Data assimilation / inverse modeling in R
1. The Details for ?help.search indicate that it searches ONLY
packages currently installed locally.
2. See Also for ?help.search includes 'RSiteSearch' to access
an on-line search of R resources.
3. 'help.search(Kalman filter)' for me just now identified only
KalmanLike(stats): Kalman Filtering. However, 'RSiteSearch(Kalman
filter)' produced 80 documents, and 'RSiteSearch(Kalman filter,
functions)' produced 18.
4. RSiteSearch(assimilation) produced only 3 hits, being your post
and two others connected to it appearently because two other posters
just replied to your post, changing subject and content, not realizing
that the RSiteSearch software somehow keeps those connections.
5. RSiteSearch(inverse model) produce 723 hits, so that's not much
help either.
6. Have you read Venables and Ripley (2002) Modern Applied Statists
with S (Springer), especially the chapter on time series? I highly
recommend this book and this chapter in particular. Also, Have you
worked through the vignettes associated with the zoo package? If no,
you might find that quite useful. [Are you aware that
edit(vignette(...)) will provide a script file with the R code discussed
in the vignette, which can be viewed in Adobe Acrobat while you are
working throught the examples line by line, modifying them, etc.? I've
found this to be very useful. If you use XEmacs, edit(vignette(...))
may not work. Instead, try Stangle(vignette(...)$file). This will copy
the R code to a file in the working directory, which you can then open.]
Also, the dse bundle does some Kalman filtering and has vignettes. =
7. I think I have some familiarity with a fairly broad range of
statistical methods, but I can't guess what you might mean by data
assimilation / inverse modeling, apart from your use of the term
Kalman filter. PLEASE do read the posting guide!
www.R-project.org/posting-guide.html, then prepare another post
preferably including a small, self contained example to illustrate
something you tried, explaining why that did not quite do what you want.
hope this helps,
spencer graves
Lapointe, Pierre wrote:
Hello,
I'm trying to find out if something has been written in R regarding data
assimilation and inverse modeling.
These searches do not return anything that look like Kalman filter
variations (EK, SEEK, ROEK, etc.)
help.search(assimilation)
help.search(inverse model)
Regards,
**
AVIS DE NON-RESPONSABILITE: Ce document transmis par courrie...{{dropped}}
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Random Coefficients using coxme
Have you tried contacting the authors and / or maintainer for the 'kinship' package? Name and email address should appear with help(package=kinship); this does NOT work for me, because I don't have kinship installed. From www.r-project.org - CRAN - (select a mirror) - packages, I learned that kinship was developed by Beth Atkinson for pedigree functions [and] Terry Therneau for all other functions and maintained by Jing hua Zhao. This provided email addresses for Atkinson and Therneau but hot for Zhao. If you need more help on this, have you considered www.bioconductor? This looks like it might be closer to their interests. hope this helps. spencer graves Powers, Mark wrote: Hello, I was hoping someone could answer a question for me that may either be statistical or script related. I don't come from a statistics background, so I am not positive if I am using the correct nomenclature or even the correct procedure. Is it possible to model random coefficients in a mixed effects cox-regression using coxme from the Kinship package? For example, using lmer from the lme4 package, I can model V1 and V2 as a fixed random coefficient: Mod1=lmer(y ~ V1 + V2 + V3 + V4 + (1 + V1 + V2|GROUP)) Can I do something like this, though correctly without getting the error (Error in max(kindex) : object kindex not found)? Mod2=coxme(Surv(YDAYS, Y) ~ V1 + V2 + V3 + V4, data=ds1, random=~1 + V1 + V2|GROUP The GROUP variable is a census tract, while V1 V2 are individual-level characteristics of people. Thanks in advance. Mark Powers [EMAIL PROTECTED] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Problem with help() respectively ?
Hi, I have a problem with my the R help system. I have been searching through the help archives but I can't find anything about it and I don't know how to specify my search better. When I type for instance ?hist, just the next command line prompt shows but nothing actually happens. There is no help window coming up. I already tried reinstalling R (2.2.1, windows XP) but that didn't change anything. I am pretty sure the help has been working before on this machine. Has anybody an idea how to recover the help system? Which help format do you use? What about help(help, chmhelp=FALSE, htmlhelp=FALSE, offline=FALSE) Does this work fo you? Uwe Ligges Yes! Thanks, that works. And now it works just fine again... I have no idea what changed... Thanks for your help, Werner __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] small sample size confidence interval by bootstrap
On Friday 31 March 2006 18:21, Urania Sun wrote: Hi, All: I only have 4 samples. I wish to get a confidence interval around the mean. Is it reasonable? If not, is there a way to compute a confidence interval for such small sample size's mean? Many thanks, With a sample that small, it is far safer to simply consider them as four examples and leave it at that. In a population where there is little variation (say an archaeological projectile point type with a nech width that varies between 3 and 5 mm), the examples are likely to be close to typical, and the difference isn't really llikely to be important anyway. However, in a population with considerable variation (for example height in humans) you can see that trying to make any generalizations from 4 examples is going to be more likely to be misleading than anything else. If your sample of four is your entire population, you have all the information possible through simple measurements. But, if the population were 100 the number of possible samples of size 4 is, as Gnumeric assures me, about 4 x 10^306, which, to put it scientifically, is a whole bunch. It'is better not to generalize from small samples. JD __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Sourcing data files into different array fields
Hi, I make a number of experiments with a software which spits out one .R file of data objects to be sourced for each experiment. Now, the files repeat the data object names but ultimately I also want to perform comparision statistics between several objects. Now my question is, is there a slick way to wrap those data objects while sourcing each experiment file so that for instance I get an array m[] for each experiment and can access the objects by using m[1].object1 and so on? I am thankful for any suggestion! Werner __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] loglikelihood and lmer
On 3/31/06, Marco Geraci [EMAIL PROTECTED] wrote: Dear R users, I am estimating Poisson mixed models using glmmPQL (MASS) and lmer (lme4). We know that glmmPQL do not provide the correct loglikelihood for such models (it gives the loglike of a 'pseudo' or working linear mixed model). I would like to know how the loglike is calculated by lmer. Good point. The person who wrote the documentation (i.e. I) should have mentioned that. With lmer, one can fit a generalized linear mixed model using PQL or by optimizing the Laplacian approximation to the deviance directly. Even when you use PQL, which is the default method, the Laplace approximation is evaluated at the converged parameter estimates. This is the value of the loglikelihood that is reported. I am reconsidering having PQL as the default method for fitting generalized linear mixed models with lmer. I would appreciate it if you and others who do fit such models could tell me if it is noticeably slower or less reliable to do direct optimization of the Laplace approximaiton. That is, if you use the combination of optional arguments method = Laplace and control = list(usePQL = FALSE) does the fit take substantially longer? On your example I get system.time(fit1.lmer - lmer(z ~ X-1 + (1|id), family=poisson)) [1] 0.48 0.01 0.54 0.00 0.00 system.time(fit2.lmer - lmer(z ~ X-1 + (1|id), family=poisson, method = Laplace, control = list(usePQL = FALSE))) [1] 0.61 0.00 0.62 0.00 0.00 fit1.lmer Generalized linear mixed model fit using PQL Formula: z ~ X - 1 + (1 | id) Family: poisson(log link) AIC BIClogLik deviance 922.2406 934.8844 -458.1203 916.2406 Random effects: Groups NameVariance Std.Dev. id (Intercept) 0.82446 0.908 number of obs: 500, groups: id, 100 Estimated scale (compare to 1) 1.021129 Fixed effects: Estimate Std. Error z value Pr(|z|) X1 1.003639 0.098373 10.202 2.2e-16 *** X2 2.075037 0.052099 39.829 2.2e-16 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Correlation of Fixed Effects: X1 X2 -0.337 fit2.lmer Generalized linear mixed model fit using Laplace Formula: z ~ X - 1 + (1 | id) Family: poisson(log link) AIC BIC logLik deviance 921.958 934.6019 -457.979 915.958 Random effects: Groups NameVariance Std.Dev. id (Intercept) 0.8895 0.94313 number of obs: 500, groups: id, 100 Estimated scale (compare to 1) 0.04472136 Fixed effects: Estimate Std. Error z value Pr(|z|) X1 0.985721 0.101645 9.698 2.2e-16 *** X2 2.075060 0.052114 39.818 2.2e-16 *** --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 Correlation of Fixed Effects: X1 X2 -0.326 The only unexpected part of that output is the estimated scale which is wrong (well, it is never calculated in this case and consequently should not be displayed). A minor question is: why do glmmPQL and lmer give different degrees-of-freedom for the same estimated model? Does glmmPQL consider the scale parameter 'phi' as a degree of freedom? I believe that is the reason. The class of an object fit by glmmPQL is class(fm1) [1] glmmPQL lme and the only method I could find for the glmmPQL class is methods(class = glmmPQL) [1] predict.glmmPQL* Non-visible functions are asterisked Generic functions other than predict will choose the method for the lme class of linear mixed effects models (or, if there isn't an lme method, the default method). The lme methods defined in the nlme package are appropriate for linear mixed effects models (which is what Jose and I wrote them for) and typically are not appropriate for a generalized linear mixed model. A toy example set.seed(100) m - 5 n - 100 N - n*m X - cbind(1,runif(N)) Z - kronecker(diag(n),rep(1,m)) z - rpois(N, exp(X%*%matrix(c(1,2)) + Z%*%matrix(rnorm(n id - rep(1:n,each=m) fit.glmm - glmmPQL(z ~ X-1, random = ~1|id, family=poisson,verbose=F) fit.lmer - lmer(z ~ X-1 + (1|id), family=poisson,verbose=F) logLik(fit.glmm) 'log Lik.' -386.4373 (df=4) logLik(fit.lmer) 'log Lik.' -458.1203 (df=3) Thanks, Marco sessionInfo() R version 2.2.1, 2005-12-20, i386-pc-mingw32 attached base packages: [1] methods stats graphics grDevices utils [6] datasets base other attached packages: mvtnormSemiParcluster lme4lattice Matrix 0.7-21.0-1 1.10.4 0.995-2 0.12-11 0.995-5 nlme MASS 3.1-68.1 7.2-24 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Using vectorization instead of for loop for performing a calculation efficiently
I am trying to write an efficient function that will do the following: Given an nxm matrix, 10 rows (observations) by 10 columns (samples) for each row, test of all values in the row are greater than a value k If all values are greater than k, then set all values to NA (or something), Return an nxm matrix with the modified rows. If I do this with a matrix of 20,000 rows, I will be waiting until Christmas for it to finish: For rows in Matrix: if rows filter set all elements in rows to NA (or something) else do nothing Return the matrix with the modified rows I dont know how to code this properly. The following: If (sum(ifelse(nvectorfilter,1,0) == 0 ) ) Tells me if any row has at least 1 value above the filter. How do I get rid of the 'outer' loop? Peter - Peter Wilkinson Senior Bioinformatician / Programmer-Analyst National Immune Monitoring Laboratory tel: (514)-343-7876 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Using vectorization instead of for loop for performing a calculation efficiently
Try this: set.seed(1) mat - matrix(rnorm(2 * 10), 2) system.time(mat2 - replace(mat, rowSums(mat 0) == 10, NA)) [1] 0.04 0.01 0.05 NA NA R.version.string # Windows XP [1] R version 2.2.1, 2005-12-20 On 4/1/06, Peter Wilkinson [EMAIL PROTECTED] wrote: I am trying to write an efficient function that will do the following: Given an nxm matrix, 10 rows (observations) by 10 columns (samples) for each row, test of all values in the row are greater than a value k If all values are greater than k, then set all values to NA (or something), Return an nxm matrix with the modified rows. If I do this with a matrix of 20,000 rows, I will be waiting until Christmas for it to finish: For rows in Matrix: if rows filter set all elements in rows to NA (or something) else do nothing Return the matrix with the modified rows I don't know how to code this properly. The following: If (sum(ifelse(nvectorfilter,1,0) == 0 ) ) Tells me if any row has at least 1 value above the filter. How do I get rid of the 'outer' loop? Peter - Peter Wilkinson Senior Bioinformatician / Programmer-Analyst National Immune Monitoring Laboratory tel: (514)-343-7876 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Sourcing data files into different array fields
Use the local= argument of source.
Here we createa a list, L, whose ith component is an environment,
viz. the environment of the instance of f that did the sourcing,
containing the sourced objects. First we create some test
files: /a1 and /a2 and then we create L. The components
of L can be referred to as L[[1]], L[[2]] or using filenames
L[[a1]], L[[a2]] and the components can be extracted
via L[[a1]]$a, etc.
# create test files
files - paste(/a, 1:2, sep = )
for(i in seq(files)) cat(a -, i, \n, file = files[i])
# read them back
f - function(.x) { source(.x, local = TRUE); environment() }
L - sapply(files, f, simplify = FALSE)
ls(L[[1]]) # a
L[[1]]$a # 1
L[[/a2]][[a]] # 2
On 4/1/06, Werner Wernersen [EMAIL PROTECTED] wrote:
Hi,
I make a number of experiments with a software which
spits out one .R file of data objects to be sourced
for each experiment. Now, the files repeat the data
object names but ultimately I also want to perform
comparision statistics between several objects.
Now my question is, is there a slick way to wrap
those data objects while sourcing each experiment file
so that for instance I get an array m[] for each
experiment and can access the objects by using
m[1].object1 and so on?
I am thankful for any suggestion!
Werner
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] VARIANCE across each ROW
I have a very large matrix. I would like to display the variance across each row. In other words, I want to output a vector containing the values of variance across row. When I use the function var(), it seems to give me the variability of each column. Any ideas?? [[alternative HTML version deleted]] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Unbalanced Manova
What have you tried? I just did help.searhc(manova), which led me to manova and summary.manova, which contained a balanced example, which I unbalanced as follows: tear - c(6.5, 6.2, 5.8, 6.5, 6.5, 6.9, 7.2, 6.9, 6.1, 6.3, 6.7, 6.6, 7.2, 7.1, 6.8, 7.1, 7.0, 7.2, 7.5, 7.6) gloss - c(9.5, 9.9, 9.6, 9.6, 9.2, 9.1, 10.0, 9.9, 9.5, 9.4, 9.1, 9.3, 8.3, 8.4, 8.5, 9.2, 8.8, 9.7, 10.1, 9.2) opacity - c(4.4, 6.4, 3.0, 4.1, 0.8, 5.7, 2.0, 3.9, 1.9, 5.7, 2.8, 4.1, 3.8, 1.6, 3.4, 8.4, 5.2, 6.9, 2.7, 1.9) rate - factor(gl(2,10), labels=c(Low, High)) additive - factor(gl(2, 5, len=20), labels=c(Low, High)) DF - data.frame(tear, gloss, opacity, rate, additive) fit - manova(cbind(tear, gloss, opacity) ~ rate * additive, DF) f18 - manova(cbind(tear, gloss, opacity) ~ rate * additive, DF[1:18,]) summary(fit, test=Wilks) # ANOVA table of Wilks' lambda summary(fit, test=Wilks) # ANOVA table of Wilks' lambda Df Wilks approx F num Df den Df Pr(F) rate 1 0.3819 7.5543 3 14 0.003034 ** additive 1 0.5230 4.2556 3 14 0.024745 * rate:additive 1 0.7771 1.3385 3 14 0.301782 Residuals 16 --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 summary(f18, test=Wilks) # ANOVA table of Wilks' lambda Df Wilks approx F num Df den Df Pr(F) rate 1 0.3743 6.6880 3 12 0.006637 ** additive 1 0.5825 2.8671 3 12 0.080857 . rate:additive 1 0.7012 1.7047 3 12 0.218966 Residuals 14 --- Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 If you would like further help from this listserve, PLEASE do read the posting guide! www.R-project.org/posting-guide.html and provide a self-contained example of what you tried, explaining why it does not meet your needs. hope this helps, spencer graves Naiara S. Pinto wrote: Dear all, I need to do a Manova but I have an unbalanced design. I have morphological measurements similar to the iris dataset, but I don't have the same number of measurements for all species. Does anyone know a procedure to do Manova with this kind of input in R? Thank you very much, Naiara. Naiara S. Pinto Ecology, Evolution and Behavior 1 University Station A6700 Austin, TX, 78712 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] VARIANCE across each ROW
?apply apply(data,1,var) On Sat, 01 Apr 2006 21:27:56 -0500, mark salsburg [EMAIL PROTECTED] wrote: I have a very large matrix. I would like to display the variance across each row. In other words, I want to output a vector containing the values of variance across row. When I use the function var(), it seems to give me the variability of each column. Any ideas?? [[alternative HTML version deleted]] __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html -- Using Opera's revolutionary e-mail client: http://www.opera.com/mail/ __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
[R] Rattle: A simple R/Gnome GUI for Data Mining
An oft heard issue with R is the learning curve. Yet R is a very powerful language for data mining, if only one persists with it. Attempting to provide a fast track for anyone to learning R through a GUI, without limiting the user to just the GUI, I've pulled together some basic R functionality for typical data mining into a GUI written in R using RGtk2 (need the RGtk2 package from http://www.ggobi.org/rgtk2/). This differs from other R GUIs in that it is specifically for data mining type tasks and maps to how we typically proceed through a project. The GUI is simple and basic, covering the common tasks of loading a CSV file, selecting Variables, sampling the data, summarising the data, clustering, model building (decision tree, regression, random forrest, boosting, svm), and evaluation (confusion table, ROC, Risk Chart). Everything is logged to a Log window and direct cut-n-paste from there to the R Console should work. This provides tuition or a reminder of how to do things. I've not made it into a formal R package yet but as time permits will do so. The package has been available and in use in a number of data mining projects for a couple of months now, and is under continuing evolution. It works well for what it was designed for (basically binary classification) but should handle anything thrown at it (at least gracefully informing you if it can not). It is freely available (GPL) from http://www.togaware.com/datamining/rattle.html Comments, suggestions, bugs, and code are always welome. Regards, Graham Williams __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] matching a given sd range
Date: Fri, 31 Mar 2006 14:44:05 -0800 (PST)
From: Fred J. [EMAIL PROTECTED]
given a numeric array (a sequence of reals), I am interested in
finding the subsets of sequences (each with start and end index) which match
a given sd range.
I read the docs on match and which and the see also but could not come up
with a way. I could loop with a stepping window over the sequence but that
would be limited to a fixed size window, I guess I as well loop through all
window sizes from 1:length(data) but that is not elegant. any hints towards
a solution?
Presumably your array is a vector, in which case you can use
vectorisation to eliminate one of the loops. The following function
produces a matrix in upper triangular form for which res[i, j] is the
variance of the sequence dat[i:j] (where dat is your array). It is
(approximately) two orders of magnitude faster than the double loop,
but doesn't produce exactly the same numbers because of numerical
rounding because it uses the formula: variance = (sum(xi^2) -
n*xbar^2)/(n - 1)
Now all you have to do is test this matrix against the range of
variances you want to isolate, and use the matching indices to provide
you with the endpoints of your subsets.
Hope this helps,
Ray Brownrigg
allvar - function(dat) {
len - length(dat)
res - numeric((len - 1)*len/2)
mym - dat[1]
myv - 0
for (j in 2:len) {
oldm - mym
mym - (((j - 1):1)*mym + dat[j])/(j:2)
myv - (((j - 2):0)*myv + ((j - 1):1)*oldm^2 + dat[j]^2 - (j:2)*mym^2)/
((j - 1):1)
res[((j - 2)*(j - 1)/2 + 1):((j - 1)*j/2)] - myv
mym -c(mym, dat[j])
myv - c(myv, 0)
}
return(res)
}
# Now test it
n - 500; dat - runif(n)
unix.time({res2 - matrix(0, n, n); res2[upper.tri(res2)] - allvar(dat)})
[1] 0.64 0.00 0.64 0.00 0.00
unix.time({res1 - matrix(0, n, n); for (j in 2:n) for (i in 1:(j - 1))
res1[i, j] - var(dat[i:j])})
[1] 72.64 13.74 86.38 0.00 0.00
max(abs(res1 - res2))
[1] 1.249001e-15
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] Unbalanced Manova
Naiara S. Pinto [EMAIL PROTECTED] writes: Dear all, I need to do a Manova but I have an unbalanced design. I have morphological measurements similar to the iris dataset, but I don't have the same number of measurements for all species. Does anyone know a procedure to do Manova with this kind of input in R? Thank you very much, If you have complete *responses*, the anova and summary methods for mlm objects (from lm(Y~) where Y is a matrix) should do it. -- O__ Peter Dalgaard Øster Farimagsgade 5, Entr.B c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~ - ([EMAIL PROTECTED]) FAX: (+45) 35327907 __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
