[R-sig-ME] Question on random effects glm interpretation

[Thierry Onkelinx]

Dear Emmanuel,

Maybe a survival analysis is more appropriate for that kind of data.

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

2015-11-12 18:12 GMT+01:00 Emmanuel Curis <emmanuel.curis op parisdescartes.fr>
:

> Dear all,
>
> With a colleague, we are discussing about the appropriateness of a
> random effects mixed effects generalized linear model (more
> specifically, logistic regression) in a given experimental situation,
> and wonder about it's correct interpretation.
>
> Short version: is it correct to interpret random effects glm, with a
> single random effect, as hierarchical models as random effects linear
> models ?
>
> Detailed version: our data consist of daily status of a set of
> patients, the status beeing « Infected » or « Not infected », during a
> variable period of time. The aim is to see what changes the infection
> probability.
>
> A proposed approach was to use a random effect logistic regression to
> evaluate this probability, with patient as a random effect.
>
> However, we have one concern with that approach: interpretating it in
> a hierarchical model idea, it seems that for a given patient, the
> model should be a binomial one, in other words that the set of
> Bernoulli variables observed each day for a given patient should be
> independent, identically distributed. But this is obviously not the
> case here: the variable is 0 until infection occurs (if it occurs),
> and 1 after.  Consequently, we fear that the probability estimated in
> the glmm will have no real meaningful interpretation.
>
> Are we right with this hierarchical interpretation of the GLMM
> (logistic) model, or is this fear not justified, and the
> interpretation of the GLMM more complex, but would lead to correct and
> interpretable estimations of infection probability?
>
> As an alternative approach, we thought about a two-states Markov
> chain, and working on transition probabilities from the « non-infected »
> to the « infected » state. It seems to model what happens more
> closely.  Is there any link between such a model and the logistic GLMM
> described above, or another kind of GLMM model?
>
> And, semantic question, is such a model also in the scope of « random
> effects » model and could be discussed here, in case help is needed, or
> is it out of the scope of this list?
>
> Thanks in advance for your opinions,
> Best regards,
>
> --
>                                 Emmanuel CURIS
>                                 emmanuel.curis op parisdescartes.fr
>
> Page WWW: http://emmanuel.curis.online.fr/index.html
>
> _______________________________________________
> R-sig-mixed-models op r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>

	[[alternative HTML version deleted]]