Search String: Display: Description: Sort:

Results:

References: [ +from:Bill.Venables@cmis.csiro.au: 93 ]

Total 93 documents matching your query.

1. Re: correlation (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Wed, 8 Jan 2003 08:44:05 +1000
optimal correlation is achieved by giving the levels of the factor variable the class means for the other variable. A simple function to do what you want would be: cor.xf <- function(x, f) { m <- ta
/archives/html/s-news/2003-01/msg00021.html (9,273 bytes)

2. Population modeller position available (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Mon, 9 Dec 2002 09:02:52 +1000
There is a population modeller position available at this laboratory for someone with strong modelling and computational skills. Full details are available at the following web site. <http://recruitm
/archives/html/s-news/2002-12/msg00072.html (6,922 bytes)

3. Re: Type I and III sum of squares (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Wed, 11 Dec 2002 18:51:59 +1000
think you have put your finger on one of the main problems with Type III sums of squares. What you are asking for (additivity and order invariance) is not possible for a non-orthogonal design. Type
/archives/html/s-news/2002-12/msg00084.html (9,669 bytes)

4. Re: modeling log transformed values in lme (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Thu, 17 Oct 2002 11:20:26 +1000
Buzz Burhans asks: [WNV] I take it you mean that after log transformation of the data the residuals appear more normally distributed, after fitting appropriate linear models. [WNV] No, there is not,
/archives/html/s-news/2002-10/msg00123.html (9,491 bytes)

5. Re: VIF Variance Inflation Factor (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Tue, 22 Oct 2002 11:54:35 +1000
Kenneth Cabrera asks: [WNV] this comes up every now and then and I suppose it has been answered dozens of times, but here is a simple version of a generic function that people might find useful (and
/archives/html/s-news/2002-10/msg00156.html (9,112 bytes)

6. Re: indexing feature (WAS indexing within a list using Splus (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Mon, 28 Oct 2002 18:02:02 +1000
Just a minor footnote to Brian's reply below. The only place where this kind of indexing does seem to be useful is in dealing with language objects. lists of lists of lists ... are not all that usefu
/archives/html/s-news/2002-10/msg00236.html (11,153 bytes)

7. Re: how to get rid of this double for loop? (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Fri, 1 Nov 2002 14:10:37 +1000
Danni Wang asks: [WNV] I presume you are assuming that the very first element of every column is not missing and that you never get two consecutive missing values in any column. You did say this is a
/archives/html/s-news/2002-10/msg00279.html (7,938 bytes)

8. Re: fit the MLE of gamma dist (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Sun, 29 Sep 2002 11:25:12 +1000
take a bigger sample: rate = 5) > fm <- glm(y ~ 1, Gamma(link = identity)) library(MASS) gamma.shape(fm)$alpha > rate <- shape/coef(fm) > c(shape = shape, rate = as(rate, "vector")) shape rate 4.016
/archives/html/s-news/2002-09/msg00204.html (11,597 bytes)

9. Re: fit the MLE of gamma dist (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Tue, 1 Oct 2002 09:18:15 +1000
I gone on a bit further, this would have been the next possibility, of course. I'm inclined to think, though, that optim( ) provides a better general optimiser than nlminb( ) and I thought people sh
/archives/html/s-news/2002-09/msg00219.html (15,225 bytes)

10. Re: Stepwise and significance (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Fri, 2 Aug 2002 00:34:07 +1000
Theodore Kypreos asks: --Original Message-- From: Theodore Kypreos [mailto:kypreost@yahoo.gr] Sent: Friday, August 02, 2002 12:09 AM To: s-news@lists.biostat.wustl.edu Subject: [S] Stepwise and signi
/archives/html/s-news/2002-08/msg00003.html (10,837 bytes)

11. Re: remove all data objects (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Fri, 2 Aug 2002 14:11:43 +1000
Robert Burbridge asks: [WNV] What do you mean by "data" objects? This is a highly ambiguous term and not everyone shares your view. Sorry to be picky but this is the nub of your problem. [WNV] actual
/archives/html/s-news/2002-08/msg00010.html (10,542 bytes)

12. Population dynamicist job in Queensland, Australia (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Wed, 21 Aug 2002 15:53:03 +1000
My apologies for he very short time before things close, but it's not my fault. If necessary get your hat in the ring by Friday and see to the niceties later, may I suggest. This is a vacancy in my g
/archives/html/s-news/2002-08/msg00121.html (8,670 bytes)

13. Re: query on column vector manipulation (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Thu, 22 Aug 2002 14:41:00 +1000
Moto Hasegawa asks: --Original Message-- From: motohasegawa [mailto:2116242101@jcom.home.ne.jp] Sent: Thursday, August 22, 2002 8:04 AM To: splus Subject: [S] query on column vector manipulation Dear
/archives/html/s-news/2002-08/msg00130.html (10,116 bytes)

14. Re: guiOpenView (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Fri, 12 Jul 2002 09:48:18 +1000
It is probably (almost surely, in fact) a visibility problem on the search path. Many functions are designed to work effectively at the command level and do not translate easily into functions where
/archives/html/s-news/2002-07/msg00075.html (8,038 bytes)

15. Re: Tree regression puzzlement (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Mon, 15 Jul 2002 13:23:09 +1000
David Parkhurst asks: [WNV] The problem is that you only have 16 observations. The default rpart control is for minsplit to be 20. Unless a node has at least minsplit observations a split will not be
/archives/html/s-news/2002-07/msg00087.html (9,014 bytes)

16. Re: Vectorizing the calculation of pairwise differences (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Mon, 15 Jul 2002 15:49:36 +1000
X be the matrix. Then the matrix of all pairwise differences of elements is A <- outer(as.vector(X), as.vector(X), "-") This is an n^2 x n^2 matrix but it is not symmetric - in fact it's skew-symmet
/archives/html/s-news/2002-07/msg00089.html (8,880 bytes)

17. Re: nonlinear regression: complicating the Puromycin example (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Thu, 25 Jul 2002 01:52:09 +1000
Albertus Smit asks: [WNV] There is a curiously little-known method for fitting them all in one: fm <- nls(rate ~ Vm[state] * conc/(K[state] + conc), data = Puromycin, start = list(Vm = c(200, 160, 20
/archives/html/s-news/2002-07/msg00175.html (11,376 bytes)

18. Re: What is the category of this object? (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Thu, 25 Jul 2002 13:37:59 +1000
[WNV] nope, and it wouldn't help. Rule 1: Print (at least part of) it. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 0 -1 -2 -3 -4 1 0 -1 -2 -3 2 1 0 -1 -2 3 2 1 0 -1 4 3 2 1 0 R
/archives/html/s-news/2002-07/msg00178.html (8,945 bytes)

19. Re: How to handle BIG matrices (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Wed, 31 Jul 2002 11:32:50 +1000
Your assumption that to perform principal components analysis you need the covariance matrix is not correct. You only need that if you want to do it using the usual tools, but the computation itself
/archives/html/s-news/2002-07/msg00224.html (9,223 bytes)

20. Re: Lengths of x and y must match (score: 1)
Author: Bill.Venables@CMIS.CSIRO.AU
Date: Sat, 8 Jun 2002 16:15:28 +1000
The common reason for this to happen is that you have some other version of "Year" or "Time" in your .Data directory which is masking the one you want to plot, and it is shorter or longer than the on
/archives/html/s-news/2002-06/msg00067.html (8,498 bytes)

Current List: 1 - 20
Page: [1] [2] [3] [4] [5]
This search system is powered by Namazu