fitting beta and zero mixture model containing both nested and crossed random effects

Wed, Jun 13, 2018 1:26 AM

My bad, I replied to you the first time without including the list.
Regarding your last question, perhaps the list and/or Ben could
provide a more accurate answer than me.
I'm also curious to know how glmmTMB could do that

2018-06-13 0:09 GMT+02:00 Meng Liu <liumeng at usc.edu>:

Hi Guillaume,

Thank you so much for this! I just have another question: for example if I
have random factor A and B in both logistic model part and beta model part,
then after I fit the whole model and got variance component estimation of
random effect for factor A and B for both logistic model part and beta model
model part, will there be any way to combine variance together? I.e. I can
estimate a total variance from factor A, and a total variance from factor B
(i.e. only differ by factor, not model)? Something like variance
decomposition but I believe here is more complex as this is a mixture model.

Thank you again for all your help

Best regards,

Meng

On Sun, Jun 10, 2018 at 11:03 AM, Guillaume Chaumet
<guillaumechaumet at gmail.com> wrote:

brms:
https://urldefense.proofpoint.com/v2/url?u=https-3A__cran.r-2Dproject.org_web_packages_brms_index.html&d=DwIBaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=Ij73g98b5MaGitndhxmoIw&m=Uy-z_keMG1SfZG-g8FxVqzfz-Ghl2OHun7TY7tfexwo&s=Gfi89kd1PSimpIhWBglYPuJRn3_FF_uNBGvzVDvWe4A&e=

2018-06-09 21:06 GMT+02:00 Meng Liu <liumeng at usc.edu>:

To whom it may concern,

I am trying to fit a model for a data among which the response value is
within [0,1). I am thinking about fitting the zeros as a complete
separate
category from the non-zero data, i.e. a binomial (Bernoulli) model to
"==0
vs >0" and a Beta model to the >0 responses. Also, my data contains both
nested factors and crossed factors, which means I need to add nested
random
effects and crossed random effects to both logistic model part and beta
model model. However, I didn't find any R packages can do exactly what I
want (By far I found gamlss, glmmTMB, zoib but they either can only
assume
random zero or they can only fit repeated measures/clustered data but
not
nested and crossed design). Therefore, I am wondering if any one know if
there is any available package or function can do this.

Thank you very much for your help!

Best regards

Meng

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-mixed-models at r-project.org mailing list

https://urldefense.proofpoint.com/v2/url?u=https-3A__stat.ethz.ch_mailman_listinfo_r-2Dsig-2Dmixed-2Dmodels&d=DwIBaQ&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=Ij73g98b5MaGitndhxmoIw&m=Uy-z_keMG1SfZG-g8FxVqzfz-Ghl2OHun7TY7tfexwo&s=FMNtOORgf7OlXhD5m8VHoGCnuWlt5NLqtXxalxQOhQw&e=

fitting beta and zero mixture model containing both nested and crossed random effects

Thread (4 messages)