[R-sig-ME] c++ exception with logistic glmer

[Cole, Tim]

Thanks Emmanuel. I haven't come across msm.

In my previous post I forgot to respond to David's comment about growth curve modelling. As it happens I have already analysed and published the data that way, using my SITAR model which includes a subject-specific random effect on the age scale. This is exactly analogous to the random effect I am trying to fit with glmer, and hence is another reason why I prefer that approach.

Best wishes,
Tim
---
Tim.Cole at ucl.ac.uk<mailto:Tim.Cole at ich.ucl.ac.uk> Phone +44(0)20 7905 2666 Fax +44(0)20 7905 2381
Population Policy and Practice Programme
UCL Institute of Child Health, London WC1N 1EH, UK

From: Emmanuel Curis <emmanuel.curis at parisdescartes.fr<mailto:emmanuel.curis at parisdescartes.fr>>
Date: Thursday, 8 January 2015 11:46
To: Tim Cole <tim.cole at ucl.ac.uk<mailto:tim.cole at ucl.ac.uk>>
Cc: David Duffy <David.Duffy at qimr.edu.au<mailto:David.Duffy at qimr.edu.au>>, "r-sig-mixed-models at r-project.org<mailto:r-sig-mixed-models at r-project.org>" <r-sig-mixed-models at r-project.org<mailto:r-sig-mixed-models at r-project.org>>
Subject: Re: [R-sig-ME] c++ exception with logistic glmer

Tim,

Did you looked at the msm package which, IIRC, allows to implement a
hidden markov model with hidden states the real state (with only 0 ->
1 transition allowed) and allowed states the observed states (which
can then have 1 -> 0 sequences), with transition matrix probabilities
and measurement errors matrixes? This should handle the measurement
errors problems you mention, also handling various censoring.

However, I am not sure it will easily allow to estimate a
between-subject SD of the time to maturity, you should check the
package doc and the underlying model to see if one can access each
individual transition time to have your answer... I am unfortunately
not familiar enough with it to confirm or not that.

On Thu, Jan 08, 2015 at 11:33:42AM +0000, Cole, Tim wrote:
« David,
«
« I agree that the model can be fitted as an interval-censored time-to-event (actually it's a mix of left, right and interval-censored), but that does not make my approach wrong. In fact it is better in two respects.
«
« First, bone maturity score is subject to measurement error, so it's possible (though rare) to transition from 0 to 1 and then back to 0. Survival analysis treats such points as missing, when they are a valid representation of measurement error.
«
« Second, I am interested in the between-subject SD of the time to maturity, obtained by dividing the random intercept SD by the age coefficient. To my knowledge this is not available from the survival analysis.
«
« Whether or not you are persuaded by these arguments, the fact remains that my two questions about glmer remain valid - its behavour is odd.
«
« Best wishes,
« Tim
« ---
« Tim.Cole at ucl.ac.uk<mailto:Tim.Cole at ucl.ac.uk><mailto:Tim.Cole at ich.ucl.ac.uk> Phone +44(0)20 7905 2666 Fax +44(0)20 7905 2381
« Population Policy and Practice Programme
« UCL Institute of Child Health, London WC1N 1EH, UK
«
« From: David Duffy <David.Duffy at qimr.edu.au<mailto:David.Duffy at qimr.edu.au><mailto:David.Duffy at qimr.edu.au>>
« Date: Thursday, 8 January 2015 05:32
« To: Tim Cole <tim.cole at ucl.ac.uk<mailto:tim.cole at ucl.ac.uk><mailto:tim.cole at ucl.ac.uk>>
« Cc: "r-sig-mixed-models at r-project.org<mailto:r-sig-mixed-models at r-project.org><mailto:r-sig-mixed-models at r-project.org>" <r-sig-mixed-models at r-project.org<mailto:r-sig-mixed-models at r-project.org><mailto:r-sig-mixed-models at r-project.org>>
« Subject: Re: [R-sig-ME] c++ exception with logistic glmer
«
« On Tue, 6 Jan 2015, Cole, Tim wrote:
«
« I'm returning to a thread I started a year ago. My logistic glmer model
« [...]B
« and Jonathan French and Ben suggested an alternative time-to-event
« analysis instead of the logistic.
«
« As now described, your problem is clearly interval censored time-to-event,
« and your logistic model is just not the right approach - bone maturity is
« an irreversible state (consider what happens to your age regression
« coefficient if you add in more ages before or after maturity).  Either do
« the survival analysis, which gives you the median age at maturity, or fit
« a (nonlinear) growth model to maturity score.
«
« | David Duffy (MBBS PhD)
« | email: David.Duffy at qimrberghofer.edu.au<mailto:David.Duffy at qimrberghofer.edu.au><mailto:David.Duffy at qimrberghofer.edu.au>  ph: INT+61+7+3362-0217 fax: -0101
« | Genetic Epidemiology, QIMR Berghofer Institute of Medical Research
« | 300 Herston Rd, Brisbane, Queensland 4006, Australia  GPG 4D0B994A
«
«
« [[alternative HTML version deleted]]
«
« _______________________________________________
« R-sig-mixed-models at r-project.org<mailto:R-sig-mixed-models at r-project.org> mailing list
« https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models

--
                                Emmanuel CURIS
                                emmanuel.curis at parisdescartes.fr<mailto:emmanuel.curis at parisdescartes.fr>

Page WWW: http://emmanuel.curis.online.fr/index.html


	[[alternative HTML version deleted]]