[R-sig-ME] Repeated Measures Design With lme Function

[Thierry Onkelinx]

Dear Telli,

Your nlme syntax assumes nested random effects. Crossed random effects
are hard (but doable) with nlme. They are straightforward with lme4.
So go for lme4, useless you really need something that nlme provides
but lme4 doesn't.

Best regards,

Thierry



ir. Thierry Onkelinx
Statisticus / Statistician

Vlaamse Overheid / Government of Flanders
INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE
AND FOREST
Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance
thierry.onkelinx at inbo.be
Havenlaan 88 bus 73, 1000 Brussel
www.inbo.be

///////////////////////////////////////////////////////////////////////////////////////////
To call in the statistician after the experiment is done may be no
more than asking him to perform a post-mortem examination: he may be
able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does
not ensure that a reasonable answer can be extracted from a given body
of data. ~ John Tukey
///////////////////////////////////////////////////////////////////////////////////////////




2018-02-12 18:53 GMT+01:00 Telli Davoodi <telli at bu.edu>:
> Dear Thierry,
>
> Thank you for the reply. Just to make sure, after I did a little more
> research on my design, I concluded that in my design, Question and Category
> are crossed (maybe you have already figured this our from my original
> explanation). So, with this in mind, does the syntax you suggest still make
> sense with the lme function? Or do I have to run my model with lmer?
>
> Thanks again,
> Telli
>
> On Mon, Feb 12, 2018 at 11:18 AM, Thierry Onkelinx
> <thierry.onkelinx at inbo.be> wrote:
>>
>> Dear Telli,
>>
>> When you nest question in category, you assume that you have 25 (= 5 x
>> 5) different question which are grouped into 5 categories.
>>
>> In lme4 notation the random effects (1|SubjectID/Category/Question)
>> translate more verbose into (1|SubjectID) + (1|SubjectID:Category) +
>> (1|SubjectID:Category:Question). This make it more clear that you have
>> 3 random intercepts: one per SubjectID, one per combination of
>> SubjectID and Category and one per combination of SubjectID, Category
>> and Question. That last random intercept has only one observation per
>> level and thus doesn't make sense. So I'd go for lme(Essen ~
>> AgeinYears * Category * Question, random = ~1|SubjectID/Category) Note
>> that I've rearraged the order of the fixed effects. This is yield a
>> different parametrisation which seems a bit more appropriate for this
>> case.
>>
>> Best regards,
>>
>> ir. Thierry Onkelinx
>> Statisticus / Statistician
>>
>> Vlaamse Overheid / Government of Flanders
>> INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE
>> AND FOREST
>> Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance
>> thierry.onkelinx at inbo.be
>> Havenlaan 88 bus 73, 1000 Brussel
>> www.inbo.be
>>
>>
>> ///////////////////////////////////////////////////////////////////////////////////////////
>> To call in the statistician after the experiment is done may be no
>> more than asking him to perform a post-mortem examination: he may be
>> able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher
>> The plural of anecdote is not data. ~ Roger Brinner
>> The combination of some data and an aching desire for an answer does
>> not ensure that a reasonable answer can be extracted from a given body
>> of data. ~ John Tukey
>>
>> ///////////////////////////////////////////////////////////////////////////////////////////
>>
>>
>>
>>
>> 2018-02-10 21:59 GMT+01:00 Telli Davoodi <telli at bu.edu>:
>> > Hi all,
>> >
>> > I'm having a hard time defining a model with my repeated measures design
>> > (explained below). This is an experimental design and it is fully
>> > balanced.
>> > I would really appreciate it if you have any feedback.
>> >
>> > Thanks,
>> > Telli
>> >
>> > Here's the design of my experiment: I have 72 children (Subject ID)
>> > answer
>> > five questions (Question) about five different categories (Category).
>> > Basically, the same five questions repeat for each category. So, in long
>> > format, every participant has 25 rows (five for each level of category
>> > and
>> > within each level of category, question has five levels).
>> >
>> > Now, I want to fit a mixed-effects regression model on this data, using
>> > the
>> > lme function from the nmle package, but I'm not sure how to account for
>> > the
>> > fact that Category repeats with Subject ID and Question repeats with
>> > Category. This is what I have so far, but I'm not sure if I'm specifying
>> > the random part of the model correctly:
>> >
>> > summary(Qs <- lme(Essen ~ Question * AgeinYears * Category,
>> > random = ~1|SubjectID/Category/Question, data = Q, na.action = na.omit))
>> >
>> > I just want to make sure that I am allowing Category to vary within
>> > SubjectID and Question to vary within Category.
>> >
>> > Also, is it correct to say that Category is "nested" under Subject ID
>> > and
>> > Question is nested under Category?
>> >
>> >         [[alternative HTML version deleted]]
>> >
>> > _______________________________________________
>> > R-sig-mixed-models at r-project.org mailing list
>> > https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>