[R-sig-ME] Time-varying random effects

[Thierry Onkelinx]

Dear Marc,

I would use al least (1|Sow/Pig) + (1|Blocking). Each pig will have 3
different values for Blocking, one for each stage. At the litter stage this
will coincide with Sow. However Sow rather indicates the overall genetic
effect, where Blocking at the farrowing stage indicates the additional
effect of the circumstances at that point in time.

A (1|Pen) effect is only relevant in case the same pens are used to house
multiple Blocking during the study.

Some example data describing the design would be useful.

Best regards,

ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature and
Forest
team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
Kliniekstraat 25
1070 Anderlecht
Belgium

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

2016-11-25 8:05 GMT+01:00 Marc Jacobs <marc.jacobs012 op gmail.com>:

> Hi all,
>
> thnx tor the replies but in the answers I have not found what I was
> looking for.
>
> The pigs are separated from their litter in the nursery phase being placed
> in pens based on a blocking factor (Bodyweight). This happens again in the
> growth-finish fase. Thus yes, they are moved around at least two times, all
> of them.
>
> Hence, although the genetic similarity remains across the entire study
> (pigs nested in sows), there are crossed effects with blocks, rooms, and
> pen, because it changes. Since pigs are social animals, the pen effect
> should matter and hence should be taken into account. The Blocking effect
> speaks for itself I think.
>
> Normally, this data set would be analyzed three times - once for the
> farrowing phase, once for the nursery phase, and once for the growth finish
> fase. This way, you have no time-varying RANDOM effects, but I want to
> model the entire growth curve, whilst taking into account random factors
> that change over time.
>
> Thank you,
>
> Marc
>
> 2016-11-24 15:43 GMT+01:00 Thierry Onkelinx <thierry.onkelinx op inbo.be>:
>
>> Hi Mark,
>>
>> I have some questions on the design.
>> - Can you identify the individual pigs in the data?
>> - How is the grouping of the pigs? Is it constant (e.g. all pigs from the
>> same litter stay together)? Or does the grouping changes over time?
>> - Do expect any effect of the pens itself? Or are the pens rather a just
>> group of pigs.
>>
>> Best regards,
>>
>> Thierry
>>
>> ir. Thierry Onkelinx
>> Instituut voor natuur- en bosonderzoek / Research Institute for Nature
>> and Forest
>> team Biometrie & Kwaliteitszorg / team Biometrics & Quality Assurance
>> Kliniekstraat 25
>> 1070 Anderlecht
>> Belgium
>>
>> 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
>>
>> 2016-11-23 15:58 GMT+01:00 Marc Jacobs <marc.jacobs012 op gmail.com>:
>>
>>> Hi,
>>>
>>>
>>>
>>> By request of Prof. Bolker, i am posting my question here.
>>>
>>>
>>> I am currently in the process of analyzing a growth model in pigs. Due to
>>> the confidentiality of the data, I cannot add any data which is of course
>>> the preferred course, but I hope to gain some insight here. I apologize
>>> in
>>> advance if the description is unclear.
>>>
>>>
>>>
>>> The data shows growth in 300+ pigs over 168 days, measured on 11
>>> time-points. These 168 days can be divided in three separate phases:
>>> farrowing/mom (2 timepoints), nursery (4 timepoints), and growth-finish
>>> (5
>>> timepoints).
>>>
>>>
>>>
>>> During each of these phases, the animals are placed in different rooms
>>> and
>>> pens (nested in the rooms), which by definition are random factors. Also,
>>> there is a genetic dependency of pigs (litter) nested in moms, which
>>> would
>>> be a crossed effect, since the effect takes place across the entire
>>> dataset, separate from the room/pen (pigs are separated from the litter
>>> after the farrowing/mom phase).
>>>
>>>
>>>
>>> As such, from my point of view, the room/pen are now time-varying random
>>> effects. Since I wish to model the entire growth curve, I was wondering
>>> if
>>> anybody knows how to incorporate time-varying random effects?
>>>
>>>
>>>
>>> My gut feeling tells me this is quite easy, but my models do not
>>> converge.
>>>
>>>
>>>
>>> If you need more information, please let me know.
>>>
>>>
>>>
>>> Marc
>>>
>>>         [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-mixed-models op r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>
>>
>>
>

	[[alternative HTML version deleted]]