[R-sig-ME] longitudinal analysis using lmer?
Murray Jorgensen
maj at stats.waikato.ac.nz
Tue Sep 1 04:14:28 CEST 2009
Andrew Gelman at least seems to glory in the ability of random effects
models to cope with cases having meagre data. See the discussion in the
Gelman/Hell book on the Radon data and Lac Qui Parle County.
Murray Jorgensen
Emmanuel Charpentier wrote:
> Le lundi 31 août 2009 à 12:53 -0400, Robert Terwilliger a écrit :
>> Thanks everyone for all the advice.
>>
>> One question I have (maybe there will be more...... :-P ):
>> Should I exclude subjects that have only 1 or 2 data points?
>
> The question you shuld try to answer is "*Why* do they have only 1 or 2
> points ?".
>
> HTH,
>
> Emmanuel Charpentier
>
>
>> On Fri, Aug 28, 2009 at 7:07 PM, Ken Beath<ken-PJqznCQlsrTvnOemgxGiVw at public.gmane.org> wrote:
>>> On 29/08/2009, at 6:07 AM, Robert Terwilliger wrote:
>>>
>>>> One more thing........
>>>>
>>>> What i sent was only a small sample of the data, just for the purpose
>>>> of showing what kind of set we have.
>>>>
>>>> We have about 150 subjects, with starting ages between 8 and 21, with
>>>> 3-5 data points (yearly visits) per subject.
>>>>
>>> This will be fine, although it isn't as good as having a smaller number of
>>> complete series. One point is that they don't look completely linear, so a
>>> polynomial (maybe quadratic) or regression spline may be a better option.
>>> Judging by the scatter the random effect variance will probably be close to
>>> zero.
>>>
>>> Ken
>>>
>>>
>>>
>>>
>>>> Thanks,
>>>>
>>>> --
>>>> Robert Terwilliger
>>>> Physicist
>>>> Laboratory of Neurocognitive Development
>>>> Western Psychiatric Institute and Clinic
>>>> University of Pittsburgh Medical Center
>>>> Loeffler Building
>>>> 121 Meyran Avenue #114
>>>> Pittsburgh, PA 15213
>>>> 412.383.8174 - Office
>>>> 412.383.8179 - Fax
>>>> em: raterwil at gmail.com
>>>> http://www.wpic.pitt.edu/research/lncd/
>>>>
>>>> *******************************************************
>>>> Dear R mixed effects gurus,
>>>>
>>>> I have the following data below. Attached is a png graphic
>>>> representing the data.
>>>>
>>>> I would like to run the following analysis:
>>>>
>>>> signal ~ age | subject.
>>>>
>>>> For your information (not statically relevant), the "signal" variable
>>>> is from a functional MRI experiment.
>>>>
>>>> At issue is whether this analysis is valid using "lme". From the graph
>>>> (and the table below), one can see that there are five subjects.
>>>> However, each subject begins at a different age. Subject 1 begins at 8
>>>> and goes to 12, while subject 5 begins at 14. From my study of
>>>> longitudinal analysis, usually each subject begins at the same
>>>> starting point, while these data have subjects beginning at different
>>>> starting points (different ages).
>>>>
>>>> Any insight is appreciated.
>>>>
>>>> subject age signal
>>>> 1 8 0.108
>>>> 1 9 0.139
>>>> 1 10 NA
>>>> 1 11 0.151
>>>> 1 12 0.148
>>>> 2 10 0.127
>>>> 2 11 NA
>>>> 2 12 0.135
>>>> 2 13 0.146
>>>> 3 9 0.105
>>>> 3 10 0.123
>>>> 3 11 0.134
>>>> 3 12 0.151
>>>> 3 13 0.145
>>>> 4 12 0.130
>>>> 4 13 0.169
>>>> 4 14 0.146
>>>> 4 15 0.174
>>>> 5 14 0.158
>>>> 5 15 0.141
>>>> 5 16 0.178
>>>> 5 17 NA
>>>> 5 18 0.172
>>>>
>>>> _______________________________________________
>>>> R-sig-mixed-models at r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>>
>>>
>>
>>
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
--
Dr Murray Jorgensen http://www.stats.waikato.ac.nz/Staff/maj.html
Department of Statistics, University of Waikato, Hamilton, New Zealand
Email: maj at waikato.ac.nz Fax 7 838 4155
Phone +64 7 838 4773 wk Home +64 7 825 0441 Mobile 021 0200 8350
More information about the R-sig-mixed-models
mailing list