When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
Is anyone using the lmList() function in the lme4 package?
11 messages · Douglas Bates, Reinhold Kliegl, Jonathan Baron +7 more
We are using lmList() in the context of NLMM. How easy will it be to get good starting values for nlmer? Reinhold Kliegl
On 26.06.2010, at 18:34, Douglas Bates wrote:
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
I use lmList() a fair bit because it is sort of a standard approach to data analysis in psychology when you can fit the model to each subject and then test the coefficients across subjects, or look at the correlations of the coefficients with something else. HOWEVER, for years before lme4 existed I was doing exactly the same thing with a loop. Thus, lmList() is merely a convenience for people like me. So the answer is that I WOULD BE HAPPY TO LIVE WITHOUT IT. Jon P.S. I hope that, eventually, lme4a works with languageR's pvals.fnc(). Right now it doesn't, and this is probably not your problem but theirs.
Jonathan Baron, Professor of Psychology, University of Pennsylvania Home page: http://www.sas.upenn.edu/~baron
I use it - more for exploratory analysis than to get starting estimates, to decide what might be a good starting point for the random structure of (g)lmm's. It is also very convenient to explain the concepts of mixed models, e.g. with MSc students with a biological background. Indeed, I would miss the lmList function if it was removed from future versions of lme4. All the best, Renaud Douglas Bates a ?crit :
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Renaud Lancelot EDEN Project, coordinator http://www.eden-fp6project.net/ UMR CIRAD-INRA "Contr?le des maladies animales exotiques et ?mergentes" Joint research unit "Control of emerging and exotic animal diseases" CIRAD, Campus International de Baillarguet TA A-DIR / B F34398 Montpellier http://umr-cmaee.cirad.fr/ Tel. +33 4 67 59 37 17 - Fax +33 4 67 59 37 95 Secr. +33 4 67 59 37 37 - Cell. +33 6 77 52 08 69
Douglas Bates wrote:
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
Speaking as one is relatively new to mixed models and not altogether comfortable with them at times, I have found it useful to understand my data. With some data I was having trouble with (estimated correlations of 1) it was very handy for producing caterpiller plots to see how slope and intercept were related. Of course, if I'm the only one who feels this way, my amount of usage is not nearly sufficient to demand you keep lmList. :-)
Kevin E. Thorpe Biostatistician/Trialist, Knowledge Translation Program Assistant Professor, Dalla Lana School of Public Health University of Toronto email: kevin.thorpe at utoronto.ca Tel: 416.864.5776 Fax: 416.864.3016
Douglas Bates <bates at ...> writes:
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
I would vote for keeping it, too. Anything that facilitates exploring the data is generally useful. I always found the didactic use of lmList to motivate the selection of variables that might be treated as random effects, in the Pinheiro & Bates book very helpful. I'm using the lme4 version in a chapter that I'm writing currently, so have a vested interest in knowing whether it will be deprecated, as well. Ken
Ken Knoblauch Inserm U846 Stem-cell and Brain Research Institute Department of Integrative Neurosciences 18 avenue du Doyen L?pine 69500 Bron France tel: +33 (0)4 72 91 34 77 fax: +33 (0)4 72 91 34 61 portable: +33 (0)6 84 10 64 10 http://www.sbri.fr/members/kenneth-knoblauch.html
Not that I have used it too much, but I would prefer keeping it (for teaching, practical reasons...) El 26/06/2010 18:34, Douglas Bates escribi?:
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. (Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
--------------------------------------- Jos? M. Blanco Moreno Dept. de Biologia Vegetal (Bot?nica) Universitat de Barcelona Av. Diagonal 645 08028 Barcelona SPAIN --------------------------------------- phone: (+34) 934 039 863 fax: (+34) 934 112 842 e-mail: jmblanco at ub.edu
I would argue that it should be removed because it's pretty trivial to replace with a combination of split and lapply, or if you prefer, dlply from plyr. Hadley
On Saturday, June 26, 2010, Douglas Bates <bates at stat.wisc.edu> wrote:
When we wrote the nlme package we tried to be very careful about starting estimates and created the lmList function to allow evaluation of coefficients from within-group linear fits. ?It turns out that it is most useful for longitudinal data where there are enough observations on each subject to do within-subject fits; that is, for data like the sleepstudy data. ?Subsequent developments have made it easy enough to fit the mixed model reliably without needing a lot of preliminary work to get starting estimates. ?(Similarly, the groupedData class was a good idea at the time but no longer necessary.) Would anyone miss the lmList function if it was removed from future versions of lme4?
_______________________________________________ R-sig-mixed-models at r-project.org?mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Assistant Professor / Dobelman Family Junior Chair Department of Statistics / Rice University http://had.co.nz/
On Sat, Jun 26, 2010 at 5:52 PM, Hadley Wickham <hadley at rice.edu> wrote:
I would argue that it should be removed because it's pretty trivial to replace with a combination of split and lapply, or if you prefer, dlply from plyr.
Note that it supports a pool= argument.
Hi All While it is very easy to write a function to perform the tasks of lmList(), from the point of view of teaching students who are still learning to write their own functions, it sometimes helps to have a function that already does it well. I use lmList() as I teach, for similar reasons as were mentioned earlier in this thread. So I "vote" to keep it. Cheers Ian
On 26 June 2010 18:02, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
On Sat, Jun 26, 2010 at 5:52 PM, Hadley Wickham <hadley at rice.edu> wrote:
I would argue that it should be removed because it's pretty trivial to replace with a combination of split and lapply, or if you prefer, dlply from plyr.
Note that it supports a pool= argument.
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Ian Dworkin Assistant Professor Department of Zoology Program in Ecology, Evolutionary Biology & Behaviour Program in Genetics Michigan State University office (517) 432-6733 lab (517) 432-6730 idworkin at msu.edu https://www.msu.edu/~idworkin/
Thanks for all the responses, which do answer my question of whether lmList is used and should be retained. I will do so.
On Sun, Jun 27, 2010 at 10:06 AM, Ian Dworkin <idworkin at msu.edu> wrote:
Hi All ?While it is very easy to write a function to perform the tasks of lmList(), from the point of view of teaching students who are still learning to write their own functions, it sometimes helps to have a function that already does it well. I use lmList() as I teach, for similar reasons as were mentioned earlier in this thread. So I "vote" to keep it. Cheers Ian On 26 June 2010 18:02, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
On Sat, Jun 26, 2010 at 5:52 PM, Hadley Wickham <hadley at rice.edu> wrote:
I would argue that it should be removed because it's pretty trivial to replace with a combination of split and lapply, or if you prefer, dlply from plyr.
Note that it supports a pool= argument.
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
-- Ian Dworkin Assistant Professor Department of Zoology Program in Ecology, Evolutionary Biology & Behaviour Program in Genetics Michigan State University office (517) 432-6733 lab (517) 432-6730 idworkin at msu.edu https://www.msu.edu/~idworkin/
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models