[R-sig-ME] partially crossed design, longitudinal

[Cristiano Alessandro]

Hi all,

I have a longitudinal study in which I measure the outcome variables at
baseline condition (bas), then I apply a perturbation (pert) and I measure
the outcome variable twice (early and late after perturbation is applied),
and then I remove the perturbation (noPert) and I measure twice (early and
late after perturbation is applied).

I would like to use mixed models for this design, but I am a bit confused
on how to do it. I could just have a single fixed effects 'time' with
levels 1 to 5, where level 1 would be baseline, level 2 would be pert/early
and so on. I think this is not the best design though. Alternatively, I
could use a fixed effects 'condition' with levels bas, pert, noPert,
crossed with another fixed effect 'time' with levels early/late. However,
this last design has the problem that I do not have early/late for baseline
actually.

Do you have suggestion of what to do in a case like this?

Thanks a lot
Cristiano

[[alternative HTML version deleted]]

[Thierry Onkelinx]

Dear Christiano,

IMHO, the easiest solution would be to fit the model with the 5 level time
variable and then calculate the relevant post-hoc contrasts. e.g pert =
(pert early + pert late) / 2

Thinking about the analysis at the design stage of an experiment is
valuable.

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
***@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
///////////////////////////////////////////////////////////////////////////////////////////

<https://www.inbo.be>

Post by Cristiano Alessandro
Hi all,
I have a longitudinal study in which I measure the outcome variables at
baseline condition (bas), then I apply a perturbation (pert) and I measure
the outcome variable twice (early and late after perturbation is applied),
and then I remove the perturbation (noPert) and I measure twice (early and
late after perturbation is applied).
I would like to use mixed models for this design, but I am a bit confused
on how to do it. I could just have a single fixed effects 'time' with
levels 1 to 5, where level 1 would be baseline, level 2 would be pert/early
and so on. I think this is not the best design though. Alternatively, I
could use a fixed effects 'condition' with levels bas, pert, noPert,
crossed with another fixed effect 'time' with levels early/late. However,
this last design has the problem that I do not have early/late for baseline
actually.
Do you have suggestion of what to do in a case like this?
Thanks a lot
Cristiano
[[alternative HTML version deleted]]
_______________________________________________
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

[[alternative HTML version deleted]]

[Cristiano Alessandro]

Hi Thierry,

thanks for your help. While I understand the need of considering the
analysis at the design stage of an experiment, I thought this was a pretty
standard design. Like when testing for a drug, I have a baseline (before
treatment), then I measure at different time point during administration of
the drug (to see the time course of the treatment), and then at different
time point after interruption of the drug administration (to see
retaining). I would be interested to see if the drug is effective, and if
there is a 'time' effect during drug administration and interruption.

Do you have suggestions on how to design this kind of study better for the
future? Thanks!

Best
Cristiano


Do you have suggestions on how to design this study better for the future?

Post by Thierry Onkelinx
Dear Christiano,
IMHO, the easiest solution would be to fit the model with the 5 level time
variable and then calculate the relevant post-hoc contrasts. e.g pert =
(pert early + pert late) / 2
Thinking about the analysis at the design stage of an experiment is
valuable.
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
Havenlaan 88
<https://maps.google.com/?q=Havenlaan+88&entry=gmail&source=g> 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
////////////////////////////////////////////////////////////
///////////////////////////////
<https://www.inbo.be>

Post by Cristiano Alessandro
Hi all,
I have a longitudinal study in which I measure the outcome variables at
baseline condition (bas), then I apply a perturbation (pert) and I measure
the outcome variable twice (early and late after perturbation is applied),
and then I remove the perturbation (noPert) and I measure twice (early and
late after perturbation is applied).
I would like to use mixed models for this design, but I am a bit confused
on how to do it. I could just have a single fixed effects 'time' with
levels 1 to 5, where level 1 would be baseline, level 2 would be pert/early
and so on. I think this is not the best design though. Alternatively, I
could use a fixed effects 'condition' with levels bas, pert, noPert,
crossed with another fixed effect 'time' with levels early/late. However,
this last design has the problem that I do not have early/late for baseline
actually.
Do you have suggestion of what to do in a case like this?
Thanks a lot
Cristiano
[[alternative HTML version deleted]]
_______________________________________________
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

[[alternative HTML version deleted]]