Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB (Botta-Dukát Zoltán)
Hi Vasco, An approach called fractional outcome regression sounds like it might be suitable. It is advocated for variables in the range 0 to 1 (and including these endpoints) regards, James Message: 1 Date: Thu, 29 Nov 2018 15:23:32 +0100 From: =?UTF-8?Q?Botta-Duk=c3=a1t_Zolt=c3=a1n?= To: r-sig-ecology@r-project.org Subject: Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB Message-ID: Content-Type: text/plain; charset="iso-8859-2"; Format="flowed" Hi, I'm sure that binomial is unsuitable for relative cover. Binomial distribution are defined as number of successes in independent trials. I think this scheme cannot be applied to relative cover or visually estimated cover. It is important because both number of trials and probability of success influence mean and variance, thus both should have a meaning that correspond to terms in this scheme. Unfortunately, I have no experience with tweedie distribution. I am also interested in experience of others! In theory an alternative would be zero-inflated beta distribution (after rescaling percentage between zero to one interval). Do some has an experience (including its availability in R) with it? Cheers Zoltan 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: > Hi, > > I am trying to fit a GLMM on percent cover for each species using glmer: > >> str(cover) > 'data.frame': 102 obs. of 114 variables: > $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... > $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... > $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... > $ sp1 : int 0 0 0 1 0 0 1 ... > $ sp2 : int 0 0 0 0 0 3 3 ... > $ sp3 : int 0 1 0 0 1 3 3 ... > $ sp4 : int 1 3 13 3 3 3 0 ... > $ sp6 : int 0 0 0 0 0 0 0 ... > ... > $ tot : int 93 65 120 80 138 113 ... > > sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), data=cover, > family=binomial (link ="logit")) > > However, I wonder if binomial distribution can be used (proportion of > species cover from a total cover) or if I should fitted the GLMM with > glmmTMB (tweedie distribution)? > > I would greatly appreciate it if someone could help me. > > Cheers. > > Vasco Silva > > [[alternative HTML version deleted]] > > ___ > R-sig-ecology mailing list > R-sig-ecology@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology [[alternative HTML version deleted]] ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB
Hi, Have a look at the gamlss library that has three beta inflated families. Cheers Javier -Original Message- From: R-sig-ecology On Behalf Of scott.fos...@data61.csiro.au Sent: Friday, 30 November 2018 8:57 a.m. To: r-sig-ecology@r-project.org Subject: Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB Hi, I agree with Zoltan that bionimial is probably inappropriate, for the reasons he stated. I'm not sure that Tweedie is your solution though -- it is defined for non-negative real numbers. Not just those between 0 and 100%. Perhaps easiest to think of fish biomass caught in a net (can be zero, or more. Tweedie might work though, if your percentages are typically nowhere near the 100% boundary. In this case, the upper end of the support is kind of immaterial... You hope... Does glmmTMB supply a beta distribution? Zero-inflated beta? The quantile regression idea might be useful too, as Brian suggested, but I'm not sure about random effects in that case. Beta regression will also have problems with exactly 0% (or 100%) observations. It seems, to me, that you might be forced to make a decision about what is 'least wrong', rather than what is 'best'. Scott PS Vasco and Zoltan: Sorry for the reply earlier, the message to the list bounced (CSIRO has recently changed my email address). On Thu, 2018-11-29 at 16:40 +, Vasco Silva wrote: > > Thanks Zoltan. Using the glmmTMB with tweedie is the option that I can > now discern... > > Vasco > > > > Botta-Dukát Zoltán escreveu no > dia quinta, 29/11/2018 à(s) 14:33: > > > > > > > I have to correct myself :), because an important point is missing > > from this sentence: > > > > Binomial distribution are defined as number of successes in > > independent trials. > > > > correctly: > > > > Binomial distribution are defined as number of successes in FIXED > > NUMBER OF independent trials. > > > > Zoltan > > > > 2018. 11. 29. 15:23 keltezéssel, Botta-Dukát Zoltán írta: > > > > > > > > > Hi, > > > > > > I'm sure that binomial is unsuitable for relative cover. Binomial > > > distribution are defined as number of successes in independent trials. > > > I think this scheme cannot be applied to relative cover or > > > visually estimated cover. It is important because both number of > > > trials and probability of success influence mean and variance, > > > thus both should have a meaning that correspond to terms in this scheme. > > > > > > Unfortunately, I have no experience with tweedie distribution. I > > > am also interested in experience of others! In theory an > > > alternative would be zero-inflated beta distribution (after > > > rescaling percentage between zero to one interval). Do some has an > > > experience (including its availability in R) with it? > > > > > > Cheers > > > > > > Zoltan > > > > > > 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: > > > > > > > > > > > > Hi, > > > > > > > > I am trying to fit a GLMM on percent cover for each species using glmer: > > > > > > > > > > > > > > > > > > > str(cover) > > > > 'data.frame': 102 obs. of 114 variables: > > > > $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... > > > > $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... > > > > $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... > > > > $ sp1 : int 0 0 0 1 0 0 1 ... > > > > $ sp2 : int 0 0 0 0 0 3 3 ... > > > > $ sp3 : int 0 1 0 0 1 3 3 ... > > > > $ sp4 : int 1 3 13 3 3 3 0 ... > > > > $ sp6 : int 0 0 0 0 0 0 0 ... > > > > ... > > > > $ tot : int 93 65 120 80 138 113 ... > > > > > > > > sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), > > > > data=cover, family=binomial (link ="logit")) > > > > > > > > However, I wonder if binomial distribution can be used > > > > (proportion of species cover from a total cover) or if I should > > > > fitted the GLMM with glmmTMB (tweedie distribution)? > > > > > > > > I would greatly appreciate it if someone could help me. > > > > > > > > Cheers. > > > > > > > > Vasco Silva > > > > > > > > [[alternative HTML version deleted]] > > > > > > > > ___ > > > > R-sig-ecology mailing list > > > > R-sig-ecology@r-project.org > > > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > > ___ > > > R-sig-ecology mailing list > > > R-sig-ecology@r-project.org > > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > ___ > > R-sig-ecology mailing list > > R-sig-ecology@r-project.org > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > > [[alternative HTML version deleted]] > > ___ > R-sig-ecology mailing list > R-sig-ecology@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology -- Scott Foster Research Scientist Data61, CSIRO E scott.fos...@data61.csiro.au T +61 3 6232 5178 Postal address:
Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB
Hi, I agree with Zoltan that bionimial is probably inappropriate, for the reasons he stated. I'm not sure that Tweedie is your solution though -- it is defined for non-negative real numbers. Not just those between 0 and 100%. Perhaps easiest to think of fish biomass caught in a net (can be zero, or more. Tweedie might work though, if your percentages are typically nowhere near the 100% boundary. In this case, the upper end of the support is kind of immaterial... You hope... Does glmmTMB supply a beta distribution? Zero-inflated beta? The quantile regression idea might be useful too, as Brian suggested, but I'm not sure about random effects in that case. Beta regression will also have problems with exactly 0% (or 100%) observations. It seems, to me, that you might be forced to make a decision about what is 'least wrong', rather than what is 'best'. Scott PS Vasco and Zoltan: Sorry for the reply earlier, the message to the list bounced (CSIRO has recently changed my email address). On Thu, 2018-11-29 at 16:40 +, Vasco Silva wrote: > > Thanks Zoltan. Using the glmmTMB with tweedie is the option that I can now > discern... > > Vasco > > > > Botta-Dukát Zoltán escreveu no dia > quinta, 29/11/2018 à(s) 14:33: > > > > > > > I have to correct myself :), because an important point is missing from > > this sentence: > > > > Binomial distribution are defined as number of successes in independent > > trials. > > > > correctly: > > > > Binomial distribution are defined as number of successes in FIXED NUMBER > > OF independent trials. > > > > Zoltan > > > > 2018. 11. 29. 15:23 keltezéssel, Botta-Dukát Zoltán írta: > > > > > > > > > Hi, > > > > > > I'm sure that binomial is unsuitable for relative cover. Binomial > > > distribution are defined as number of successes in independent trials. > > > I think this scheme cannot be applied to relative cover or visually > > > estimated cover. It is important because both number of trials and > > > probability of success influence mean and variance, thus both should > > > have a meaning that correspond to terms in this scheme. > > > > > > Unfortunately, I have no experience with tweedie distribution. I am > > > also interested in experience of others! In theory an alternative > > > would be zero-inflated beta distribution (after rescaling percentage > > > between zero to one interval). Do some has an experience (including > > > its availability in R) with it? > > > > > > Cheers > > > > > > Zoltan > > > > > > 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: > > > > > > > > > > > > Hi, > > > > > > > > I am trying to fit a GLMM on percent cover for each species using glmer: > > > > > > > > > > > > > > > > > > > str(cover) > > > > 'data.frame': 102 obs. of 114 variables: > > > > $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... > > > > $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... > > > > $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... > > > > $ sp1 : int 0 0 0 1 0 0 1 ... > > > > $ sp2 : int 0 0 0 0 0 3 3 ... > > > > $ sp3 : int 0 1 0 0 1 3 3 ... > > > > $ sp4 : int 1 3 13 3 3 3 0 ... > > > > $ sp6 : int 0 0 0 0 0 0 0 ... > > > > ... > > > > $ tot : int 93 65 120 80 138 113 ... > > > > > > > > sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), > > > > data=cover, > > > > family=binomial (link ="logit")) > > > > > > > > However, I wonder if binomial distribution can be used (proportion of > > > > species cover from a total cover) or if I should fitted the GLMM with > > > > glmmTMB (tweedie distribution)? > > > > > > > > I would greatly appreciate it if someone could help me. > > > > > > > > Cheers. > > > > > > > > Vasco Silva > > > > > > > > [[alternative HTML version deleted]] > > > > > > > > ___ > > > > R-sig-ecology mailing list > > > > R-sig-ecology@r-project.org > > > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > > ___ > > > R-sig-ecology mailing list > > > R-sig-ecology@r-project.org > > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > ___ > > R-sig-ecology mailing list > > R-sig-ecology@r-project.org > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > > [[alternative HTML version deleted]] > > ___ > R-sig-ecology mailing list > R-sig-ecology@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology -- Scott Foster Research Scientist Data61, CSIRO E scott.fos...@data61.csiro.au T +61 3 6232 5178 Postal address: CSIRO Marine Laboratories, GPO Box 1538, Hobart TAS 7001 Street Address: CSIRO, Castray Esplanade, Hobart Tas 7001, Australia ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB
Thanks Zoltan. Using the glmmTMB with tweedie is the option that I can now discern... Vasco Botta-Dukát Zoltán escreveu no dia quinta, 29/11/2018 à(s) 14:33: > I have to correct myself :), because an important point is missing from > this sentence: > > Binomial distribution are defined as number of successes in independent > trials. > > correctly: > > Binomial distribution are defined as number of successes in FIXED NUMBER > OF independent trials. > > Zoltan > > 2018. 11. 29. 15:23 keltezéssel, Botta-Dukát Zoltán írta: > > Hi, > > > > I'm sure that binomial is unsuitable for relative cover. Binomial > > distribution are defined as number of successes in independent trials. > > I think this scheme cannot be applied to relative cover or visually > > estimated cover. It is important because both number of trials and > > probability of success influence mean and variance, thus both should > > have a meaning that correspond to terms in this scheme. > > > > Unfortunately, I have no experience with tweedie distribution. I am > > also interested in experience of others! In theory an alternative > > would be zero-inflated beta distribution (after rescaling percentage > > between zero to one interval). Do some has an experience (including > > its availability in R) with it? > > > > Cheers > > > > Zoltan > > > > 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: > >> Hi, > >> > >> I am trying to fit a GLMM on percent cover for each species using glmer: > >> > >>> str(cover) > >> 'data.frame': 102 obs. of 114 variables: > >> $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... > >> $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... > >> $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... > >> $ sp1 : int 0 0 0 1 0 0 1 ... > >> $ sp2 : int 0 0 0 0 0 3 3 ... > >> $ sp3 : int 0 1 0 0 1 3 3 ... > >> $ sp4 : int 1 3 13 3 3 3 0 ... > >> $ sp6 : int 0 0 0 0 0 0 0 ... > >> ... > >> $ tot : int 93 65 120 80 138 113 ... > >> > >> sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), > >> data=cover, > >> family=binomial (link ="logit")) > >> > >> However, I wonder if binomial distribution can be used (proportion of > >> species cover from a total cover) or if I should fitted the GLMM with > >> glmmTMB (tweedie distribution)? > >> > >> I would greatly appreciate it if someone could help me. > >> > >> Cheers. > >> > >> Vasco Silva > >> > >> [[alternative HTML version deleted]] > >> > >> ___ > >> R-sig-ecology mailing list > >> R-sig-ecology@r-project.org > >> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > > > ___ > > R-sig-ecology mailing list > > R-sig-ecology@r-project.org > > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > > ___ > R-sig-ecology mailing list > R-sig-ecology@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-ecology > [[alternative HTML version deleted]] ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB
I have to correct myself :), because an important point is missing from this sentence: Binomial distribution are defined as number of successes in independent trials. correctly: Binomial distribution are defined as number of successes in FIXED NUMBER OF independent trials. Zoltan 2018. 11. 29. 15:23 keltezéssel, Botta-Dukát Zoltán írta: Hi, I'm sure that binomial is unsuitable for relative cover. Binomial distribution are defined as number of successes in independent trials. I think this scheme cannot be applied to relative cover or visually estimated cover. It is important because both number of trials and probability of success influence mean and variance, thus both should have a meaning that correspond to terms in this scheme. Unfortunately, I have no experience with tweedie distribution. I am also interested in experience of others! In theory an alternative would be zero-inflated beta distribution (after rescaling percentage between zero to one interval). Do some has an experience (including its availability in R) with it? Cheers Zoltan 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: Hi, I am trying to fit a GLMM on percent cover for each species using glmer: str(cover) 'data.frame': 102 obs. of 114 variables: $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... $ sp1 : int 0 0 0 1 0 0 1 ... $ sp2 : int 0 0 0 0 0 3 3 ... $ sp3 : int 0 1 0 0 1 3 3 ... $ sp4 : int 1 3 13 3 3 3 0 ... $ sp6 : int 0 0 0 0 0 0 0 ... ... $ tot : int 93 65 120 80 138 113 ... sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), data=cover, family=binomial (link ="logit")) However, I wonder if binomial distribution can be used (proportion of species cover from a total cover) or if I should fitted the GLMM with glmmTMB (tweedie distribution)? I would greatly appreciate it if someone could help me. Cheers. Vasco Silva [[alternative HTML version deleted]] ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Re: [R-sig-eco] Fitting a GLMM to a percent cover data with glmer or glmmTMB
Hi, I'm sure that binomial is unsuitable for relative cover. Binomial distribution are defined as number of successes in independent trials. I think this scheme cannot be applied to relative cover or visually estimated cover. It is important because both number of trials and probability of success influence mean and variance, thus both should have a meaning that correspond to terms in this scheme. Unfortunately, I have no experience with tweedie distribution. I am also interested in experience of others! In theory an alternative would be zero-inflated beta distribution (after rescaling percentage between zero to one interval). Do some has an experience (including its availability in R) with it? Cheers Zoltan 2018. 11. 28. 20:47 keltezéssel, Vasco Silva írta: Hi, I am trying to fit a GLMM on percent cover for each species using glmer: str(cover) 'data.frame': 102 obs. of 114 variables: $ Plot : Factor w/ 10 levels "P1","P10","P2",..: 1 1 1 1 1 3 3 ... $ Sub.plot: Factor w/ 5 levels "S1","S2","S3",..: 1 2 3 4 5 1 2 ... $ Grazing : Factor w/ 2 levels "Fenced","Unfenced": 1 1 1 1 1 1 1 ... $ sp1 : int 0 0 0 1 0 0 1 ... $ sp2 : int 0 0 0 0 0 3 3 ... $ sp3 : int 0 1 0 0 1 3 3 ... $ sp4 : int 1 3 13 3 3 3 0 ... $ sp6 : int 0 0 0 0 0 0 0 ... ... $ tot : int 93 65 120 80 138 113 ... sp1.glmm <- glmer (cbind (sp1, tot- sp1) ~ Grazing + (1|Plot), data=cover, family=binomial (link ="logit")) However, I wonder if binomial distribution can be used (proportion of species cover from a total cover) or if I should fitted the GLMM with glmmTMB (tweedie distribution)? I would greatly appreciate it if someone could help me. Cheers. Vasco Silva [[alternative HTML version deleted]] ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology ___ R-sig-ecology mailing list R-sig-ecology@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
