Re: [R] lme and lmer df's and F-statistics again
Hi Peter, thanks a lot for your help. Very much appreciated. Cheers, Julia Peter Dalgaard wrote: > > Julia S. wrote: >> Hi there, >> >> thanks for your help. I did read Bates statement several times, and I am >> very glad and thankful that many statisticians spend much time on this. >> The >> problem is, as Dieter pointed it out, that many "end users" often have to >> use statistics without being able to fully understand the math behind it. >> Because if they would spend as much time on that as statisticians do, >> they >> wouldn't be able to do what they do where they use statistics for. >> And, no, I don't expect that a "simple" answer exists, but it might be >> that >> somebody had a similar problem like me before and may have a convincing >> line >> for a referee at hands. I have problems reformulating what I read here in >> my >> own words. >> >> Dieter: when you write: >> "but to use lme instead when possible" do you mean that when using lme >> the >> F-stats are correct? Because I assumed that the problem would be the same >> with lme. >> >> Julia >> > They aren't... And they can be badly wrong in some cases. > > At this stage, I think the best one can do is to get a feeling for > whether the DF would be "large" and if so, convince the referee to > accept an asymptotic chi-square test (Wald or LRT type). > > I think that the rationale for requiring authors to state the DF is not > so much that journals believe in mighty SAS, but that they want to be > able to catch completely wrong analyses, like when people compare two > groups of each 5 rats and get a denominator DF of around 100 because > there were 10 (correlated) measurements on each rat and no between-rats > variation in the model. > > As for figuring out whether or not you have large DF; if you have a > nearly balanced design. it might be worth looking into what aov() says > would be the DF for the same model with balanced data. > > (And in any case, all DF-type corrections are in a sense wrong because > they depend on 3rd and 4th moments of the Gaussian distribution, and > your data probably aren't perfectly Gaussian.) > > -- >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@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. > > -- View this message in context: http://www.nabble.com/lme-and-lmer-df%27s-and-F-statistics-again-tp19835361p19894366.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] lme and lmer df's and F-statistics again
On 9/10/2008, at 12:34 AM, Julia S. wrote: Hm. Bert Gunter wrote: that even the most technical aspects of the discipline can be made manifest to anyone with half a brain and a stat 101 course under their belt. I don't think this is something I can use in a rebuttal. The reviewer may be offended and reviewers are people one does not want to offend. In general, I disagree. This get a bit philosophical, but well. I think there are some occasions where it is important to explain complicated things in few, easy to understand sentences to laymen (even if that means loss of preciseness). That has to be done (and was done in the past) with the other examples you give (thermodynamics, Krebs cycle ect.) fairly often, especially when politics are involved (think LHC, stem cells, or, even the structure of the DNA). Even for very difficult topics this needs to be done. I think our (maybe most challenging) duty as researchers paid by tax money is also to explain our sometimes very complicated research to laymen in an easy understandable manner. Albeit it is of course not your duty to explain it to me on this list, if you are offended by my attitude. Isn't it the most normal thing to ask for an explanation when somebody doesn't understand something? I've learned that asking is a good way of learning new things. Sorry if that offended you. This is mind-bogglingly well expressed. I wish I could write like that. Congratulations. cheers, Rolf Turner ## Attention:\ This e-mail message is privileged and confid...{{dropped:9}} __ 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] lme and lmer df's and F-statistics again
Julia S. wrote: > Hi there, > > thanks for your help. I did read Bates statement several times, and I am > very glad and thankful that many statisticians spend much time on this. The > problem is, as Dieter pointed it out, that many "end users" often have to > use statistics without being able to fully understand the math behind it. > Because if they would spend as much time on that as statisticians do, they > wouldn't be able to do what they do where they use statistics for. > And, no, I don't expect that a "simple" answer exists, but it might be that > somebody had a similar problem like me before and may have a convincing line > for a referee at hands. I have problems reformulating what I read here in my > own words. > > Dieter: when you write: > "but to use lme instead when possible" do you mean that when using lme the > F-stats are correct? Because I assumed that the problem would be the same > with lme. > > Julia > They aren't... And they can be badly wrong in some cases. At this stage, I think the best one can do is to get a feeling for whether the DF would be "large" and if so, convince the referee to accept an asymptotic chi-square test (Wald or LRT type). I think that the rationale for requiring authors to state the DF is not so much that journals believe in mighty SAS, but that they want to be able to catch completely wrong analyses, like when people compare two groups of each 5 rats and get a denominator DF of around 100 because there were 10 (correlated) measurements on each rat and no between-rats variation in the model. As for figuring out whether or not you have large DF; if you have a nearly balanced design. it might be worth looking into what aov() says would be the DF for the same model with balanced data. (And in any case, all DF-type corrections are in a sense wrong because they depend on 3rd and 4th moments of the Gaussian distribution, and your data probably aren't perfectly Gaussian.) -- 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@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] lme and lmer df's and F-statistics again
Hi there, thanks for your help. I did read Bates statement several times, and I am very glad and thankful that many statisticians spend much time on this. The problem is, as Dieter pointed it out, that many "end users" often have to use statistics without being able to fully understand the math behind it. Because if they would spend as much time on that as statisticians do, they wouldn't be able to do what they do where they use statistics for. And, no, I don't expect that a "simple" answer exists, but it might be that somebody had a similar problem like me before and may have a convincing line for a referee at hands. I have problems reformulating what I read here in my own words. Dieter: when you write: "but to use lme instead when possible" do you mean that when using lme the F-stats are correct? Because I assumed that the problem would be the same with lme. Julia -- View this message in context: http://www.nabble.com/lme-and-lmer-df%27s-and-F-statistics-again-tp19835361p19876728.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] lme and lmer df's and F-statistics again
Hm. Bert Gunter wrote: > > that even the most technical > aspects of the discipline can be made manifest to anyone with half a brain > and a stat 101 course under their belt. > > I don't think this is something I can use in a rebuttal. The reviewer may be offended and reviewers are people one does not want to offend. In general, I disagree. This get a bit philosophical, but well. I think there are some occasions where it is important to explain complicated things in few, easy to understand sentences to laymen (even if that means loss of preciseness). That has to be done (and was done in the past) with the other examples you give (thermodynamics, Krebs cycle ect.) fairly often, especially when politics are involved (think LHC, stem cells, or, even the structure of the DNA). Even for very difficult topics this needs to be done. I think our (maybe most challenging) duty as researchers paid by tax money is also to explain our sometimes very complicated research to laymen in an easy understandable manner. Albeit it is of course not your duty to explain it to me on this list, if you are offended by my attitude. Isn't it the most normal thing to ask for an explanation when somebody doesn't understand something? I've learned that asking is a good way of learning new things. Sorry if that offended you. Confused, Julia Cheers, Julia -- View this message in context: http://www.nabble.com/lme-and-lmer-df%27s-and-F-statistics-again-tp19835361p19877014.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] lme and lmer df's and F-statistics again
Bert Gunter gene.com> writes: > I think we owe Doug Bates a little more respect than that! If you check my postings on the forum and on my homepage (subject: Gastric Emptying), you will find that there are few people that pay so much respect to Douglas Bates' contributions than I do. I noted though that there was an ambiguous sentences in my previous note: "Not to use lmer for gaussian models" should have been extended by "but to use lme instead when possible". Nevertheless, there is a problem that we as "end users" are left in a dangling state. Douglas Bates has made his point in a detailed statement on this list. After 20 years in applied medical statistics I am probable at 202 level, but I simply have to accept his arguments even if I do not understand all details. However, when I like the original poster have to rebut a referee's argument, it is very difficult to use a well-formulated list message to make a point in the three-capitals dominated world. Dieter __ 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] lme and lmer df's and F-statistics again
>> Well, writing on my rebuttal, I find myself being unable to explain in a >> few, easy to understand (and, at the same time, correct) sentences stating >> that it is not a good idea to report (most likely wrong) dfs and F >> statistics. Without pretending to be able to discuss the details, may I nevertheless ask WHY one should assume that an "easy to understand (and, at the same time, correct)" answer to the question exists? For example, I would not begin to presume that an "easy to understand (and, at the same time, correct" answer exists for why an electron can simultaneously have the properties of a particle (photoelectric effect) and a wave (2 slit interference patterns)-- or to the question of "why is the 2nd law of thermodynamics equivalent to information loss?" -- or to how the Krebs cycle works or the nature of Benzene rings. It has never ceased to amaze me that many "casual" (in the sense of not having training at, say, the graduate statistics level)users of statistics automatically assume that all statistical principles are fundamentally "simple" and at least easily comprehensible at a conceptual level by someone with only minimal (or no!) background in the discipline. The extreme manifestation of this is the popular view of a statistician as someone who expertly compiles and tracks baseball records! While I would readily admit that there is much that we can and should do to make our discipline more accessible and useful, I am still offended by those whose attitude is, as appears to be the case here, that even the most technical aspects of the discipline can be made manifest to anyone with half a brain and a stat 101 course under their belt. I think we owe Doug Bates a little more respect than that! Cheers, Bert Gunter __ 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] lme and lmer df's and F-statistics again
You may find this site useful: http://lme4.r-forge.r-project.org/bib/lme4bib.html On Tue, Oct 7, 2008 at 9:08 AM, Dieter Menne <[EMAIL PROTECTED]> wrote: > Julia S. gmail.com> writes: > >> Now, I did that in my article and I got a response from a reviewer that I >> additionally should give the degrees of freedom, and the F-statistics. From >> what I read here, that would be incorrect to do, and I sort of intuitively >> also understand why (at least I think I do). > ... >> Well, writing on my rebuttal, I find myself being unable to explain in a >> few, easy to understand (and, at the same time, correct) sentences stating >> that it is not a good idea to report (most likely wrong) dfs and F >> statistics. Can somebody here help me out with a correct explanation for a >> laymen? > > Feeling with you, and hoping some day this will be resolved. I am sure you > have read Douglas Bates' > > http://finzi.psych.upenn.edu/R/Rhelp02a/archive/76742.html > > but I thought this was temporary. The only workaround I have is not to use > lmer for gaussian models. > > Dieter > > __ > 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] lme and lmer df's and F-statistics again
Julia S. gmail.com> writes: > Now, I did that in my article and I got a response from a reviewer that I > additionally should give the degrees of freedom, and the F-statistics. From > what I read here, that would be incorrect to do, and I sort of intuitively > also understand why (at least I think I do). ... > Well, writing on my rebuttal, I find myself being unable to explain in a > few, easy to understand (and, at the same time, correct) sentences stating > that it is not a good idea to report (most likely wrong) dfs and F > statistics. Can somebody here help me out with a correct explanation for a > laymen? Feeling with you, and hoping some day this will be resolved. I am sure you have read Douglas Bates' http://finzi.psych.upenn.edu/R/Rhelp02a/archive/76742.html but I thought this was temporary. The only workaround I have is not to use lmer for gaussian models. Dieter __ 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.