[R-sig-ME] glmmTMB deviance enquiry

Ben Bolker bbo|ker @end|ng |rom gm@||@com
Fri Mar 10 15:19:54 CET 2023


   I'm afraid this might be getting into more generic statistical 
territory, rather than anything specific to mixed models.  If you're 
going to rank models without regard to nesting you should probably use 
AIC rather than deviance.  Alain Zuur (Highland Statistics) has a whole 
series of books on ecological statistics that cover this topic ... 
maybe Symonds et al would be useful?


Symonds, Matthew R. E., and Adnan Moussalli. 2011. “A Brief Guide to 
Model Selection, Multimodel Inference and Model Averaging in Behavioural 
Ecology Using Akaike’s Information Criterion.” Behavioral Ecology and 
Sociobiology 65 (1): 13–21. https://doi.org/10.1007/s00265-010-1037-6.


On 2023-03-08 3:56 a.m., Rion Lerm wrote:
> Thank you @bbolker using gmail.com <mailto:bbolker using gmail.com> for your 
> elaborate response.
> 
> I am using the Deviance values in this manner to compare model fit. 
> Here, I selected for the model with second to smallest Deviance values. 
> May it be presented like this as I feel it is important to display how a 
> model was selected from various models (the reason for selecting the 
> second to smallest values aren't explained here as is the details of 
> this work):
> 
> *Model predictor fixed effects*
> 
> 	
> 
> */SR.asymp/*
> 
> 	
> 
> */SES.FRic/*
> 
> 	
> 
> */SES.PD/*
> 
> //
> 
> 	
> 
> Deviance
> 
> 	
> 
> //
> 
> Tree + Shrub + Grass + Bare ground + Infrastructure + Seasonal water + 
> Permanent water
> 
> 	
> 
> 760.4 | 1,463.6
> 
> 	
> 
> 172.1 | 319.6
> 
> 	
> 
> 206.7 | 420.7
> 
> Tree + Shrub + Bare ground + Infrastructure + Seasonal water + Permanent 
> water
> 
> 	
> 
> *766.3 | 1467.7*
> 
> 	
> 
> *179.0 | 325.2*
> 
> 	
> 
> *207.3 | 421.3*
> 
> Tree + Shrub + Grass + Bare ground
> 
> 	
> 
> 777.1 | 1,520.3
> 
> 	
> 
> 179.4 | 368.7
> 
> 	
> 
> 229.3 | 441.5
> 
> Tree + Shrub + Grass + Bare ground + Infrastructure
> 
> 	
> 
> 776.8 | 1,484.3
> 
> 	
> 
> 179.0 | 323.0
> 
> 	
> 
> 217.2 | 424.2
> 
> Tree + Shrub + Grass + Bare ground + Seasonal water + Permanent water
> 
> 	
> 
> 760.5 | 1,498.4
> 
> 	
> 
> 172.3 | 363.2
> 
> 	
> 
> 216.8 | 439.9
> 
> 	
> 
> 	
> 
> 	
> 
> 
> Importantly, is there any literature I can cite on Deviance and its use? 
> If so, please supply a DOI.
> 
> Regards/Groete,
> 
> *Rion Lerm*
> 
> SAEON Ndlovu
> 
> C: 076 913 8381
> 
> W: 013 735 3536
> 
> SAEON / Scientific Services, Kruger National Park Gate, Phalaborwa, 
> South Africa
> 
> South African Environmental Observation Network
> 
> Website <https://ndlovu.saeon.ac.za/>& Location of our offices 
> <https://www.google.co.za/maps/place/SAEON+Ndlovu+Node/@-23.9447244,31.164957,18z/data=!4m5!3m4!1s0x1ec36ec96261cfb7:0xb89a27144ba0ca2a!8m2!3d-23.9450103!4d31.1652626>
> 
> 
> 
> My research profiles:
> *ResearchGate* <https://www.researchgate.net/profile/Rion-Lerm>
> *ORCID <https://orcid.org/0000-0002-9992-2093>
> *
> *Google Scholar 
> <https://scholar.google.com/citations?user=JV5d7E4AAAAJ&hl=en>*
> 
> 
> On Tue, 7 Mar 2023 at 16:32, Ben Bolker <bbolker using gmail.com 
> <mailto:bbolker using gmail.com>> wrote:
> 
>          The deviance is twice the negative log-likelihood; if the deviance
>     is D for a model with p parameters, then AIC = D + 2*p.  The
>     deviance is
>     an *unpenalized* measure of goodness of fit (badness of fit actually -
>     lower deviance means better fit).
> 
>          There's no need to report deviance in a paper if it's not
>     useful to
>     you or your readers. The deviance for a model alone is not particularly
>     useful; you can use the deviances for several nested models to compute
>     likelihood ratio rests (via anova()), or you can use deviance as the
>     basis for computing AIC or BIC.
> 
>         Hope that helps.
> 
>         Ben Bolker
> 
>     On 2023-03-07 9:08 a.m., Rion Lerm via R-sig-mixed-models wrote:
>      > I reported, in a paper, on the *Deviance* value for model fit as
>     produced
>      > by the *glmmTMB* function.
>      >
>      > Please explain how this value is computed, and what are the
>     pros/cons of
>      > this compared to e.g., AIC. I need to comprehend its use and
>     explain the
>      > how and why? to a reviewer and possibly the paper's readers.
>      >
>      > Regards and thanks in advance,
>      >
>      > *Rion Lerm*
>      >
>      > SAEON Ndlovu
>      >
>      > C: 076 913 8381
>      >
>      > W: 013 735 3536
>      >
>      > SAEON / Scientific Services, Kruger National Park Gate,
>     Phalaborwa, South
>      > Africa
>      >
>      > South African Environmental Observation Network
>      >
>      > Website <https://ndlovu.saeon.ac.za/
>     <https://ndlovu.saeon.ac.za/>> & Location of our offices
>      >
>     <https://www.google.co.za/maps/place/SAEON+Ndlovu+Node/@-23.9447244,31.164957,18z/data=!4m5!3m4!1s0x1ec36ec96261cfb7:0xb89a27144ba0ca2a!8m2!3d-23.9450103!4d31.1652626 <https://www.google.co.za/maps/place/SAEON+Ndlovu+Node/@-23.9447244,31.164957,18z/data=!4m5!3m4!1s0x1ec36ec96261cfb7:0xb89a27144ba0ca2a!8m2!3d-23.9450103!4d31.1652626>>
>      >
>      >
>      >
>      > My research profiles:
>      > *ResearchGate* <https://www.researchgate.net/profile/Rion-Lerm
>     <https://www.researchgate.net/profile/Rion-Lerm>>
>      >
>      > *ORCID <https://orcid.org/0000-0002-9992-2093
>     <https://orcid.org/0000-0002-9992-2093>>*
>      > *Google Scholar
>      > <https://scholar.google.com/citations?user=JV5d7E4AAAAJ&hl=en
>     <https://scholar.google.com/citations?user=JV5d7E4AAAAJ&hl=en>>*
>      >
> 
>     -- 
>     Dr. Benjamin Bolker
>     Professor, Mathematics & Statistics and Biology, McMaster University
>     Director, School of Computational Science and Engineering
>     (Acting) Graduate chair, Mathematics & Statistics
>       > E-mail is sent at my convenience; I don't expect replies outside of
>     working hours.
> 
>     _______________________________________________
>     R-sig-mixed-models using r-project.org
>     <mailto:R-sig-mixed-models using r-project.org> mailing list
>     https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>     <https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models>
> 
> 
> http://www.nrf.ac.za/email-legal-notice.php 
> <http://www.nrf.ac.za/email-legal-notice.php>

-- 
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
(Acting) Graduate chair, Mathematics & Statistics
 > E-mail is sent at my convenience; I don't expect replies outside of 
working hours.



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