[R-sig-ME] WG: lme4: prediction for glmer model

Meid, Andreas Andreas.Meid at med.uni-heidelberg.de
Tue Sep 2 14:47:13 CEST 2014


Dear lme4-group members,

I'm trying to make prediction plots for visualizing effects according to http://www.ats.ucla.edu/stat/r/dae/melogit.htm. I'm worried about strange prediction plots after adapting the example code to my glmer model. More precisely, I obtain differing lines for predictions and their quartiles. I wonder if something basic is wrong. 
Is there a solution to obtain CI for predictions? Maybe the simulate function of the lme4 package is right? Or do I have to apply bootstrap function? In any case, I wonder how this is to be accomplished. I attached my files so that you can take a look at them. Hopefully, you can help.

Many thanks and best wishes,

Andreas


-----Ursprüngliche Nachricht-----
Von: Ben Bolker [mailto:bbolker at gmail.com] 
Gesendet: Dienstag, 2. September 2014 14:35
An: Meid, Andreas
Cc: lme4-authors at lists.r-forge.r-project.org
Betreff: Re: lme4: prediction for glmer model

  Can you please send this to r-sig-mixed-models at r-project.org ?  (Be aware that your .RData files probably won't make it through to the list.)

  thanks
    Ben Bolker


On 14-09-02 06:06 AM, Meid, Andreas wrote:
> Dear Ben,
> 
>  
> 
> your recently helped me with another issue about my glmer model. This 
> time I'm trying to make prediction plots for visualizing effects 
> according to http://www.ats.ucla.edu/stat/r/dae/melogit.htm. I'm 
> worried about strange prediction plots after adapting the example code 
> to my glmer model. More precisely, I obtain differing lines for 
> predictions and their quartiles. I wonder if something basic is wrong.
> 
> Is there a solution to obtain CI for predictions? Maybe the simulate 
> function of the lme4 package is right? Or do I have to apply bootstrap 
> function? In any case, I wonder how this is to be accomplished. I 
> attached my files so that you can take a look at them. Hopefully, you 
> can help.
> 
>  
> 
> Many thanks and best wishes,
> 
>  
> 
> Andreas
> 
>  
> 
> 
> *Andreas Meid*
> Heidelberg University Hospital
> Dept. of Clinical Pharmacology & Pharmacoepidemiology Im Neuenheimer 
> Feld 410
> 69120 Heidelberg
> Germany
> Phone: +49(0)6221 / 56-37113, Fax: +49(0)6221 / 56-4642
> Homepage: http://www.klinikum.uni-heidelberg.de/clinpharm
> Expert system: Drug dosing in renal failurehttp://www.dosing.de 
> <http://www.dosing.de>
> 
>  
> 



More information about the R-sig-mixed-models mailing list