individual estimations

[Fernando-UppsalaUniv]

Hi! I have had problems estimating the percentage of resource use for individual animals. The average diet of individuals within a group falls quite far away from the mean of the group calculated with the MixSIAR model. Is there any particular way to calculate individual resource use with MixSIAR that is different from the standard procedure when estimating groups? Since I have informative priors, i cannot use individual as a fixed factor as stated in the manual in the example with Fatty acids given by Galloway et al.

Any information on how to assess individual data would be very much appreciated

[brianstock]

Hi Fernando,

First, there is no one "MixSIAR model" - there are lots of options you choose that determine the model, which depends on your data and what you're interested in.

How many individuals? How many tracers? Any fixed/random/continuous effects (besides individuals)? What error model are you using? An isospace plot of the source-consumer geometry would be helpful too.

Generally, categorical effects (like "individual") are either fixed or random, so no, there isn't a different way to estimate individual diets. If you have a continuous effect, need to have individual as a random effect, and then the "process error only (MixSIR)" option. Including individual as random effect estimates a variance term for how individuals deviate from the group mean, which means adding another residual error term in the model is inappropriate.

[Fernando-UppsalaUniv]

Hej,

I have two databases that I would like to analyze with different informative priors, each of them having 56 and 57 individuals, respectively. The tracers that I use are d13C and d15N, although the isotopic separation comes mainly from C (x axis in the graph). In this particular model I would just like the estimation of diet for each individual in order to correlate those values with other continuous variables. However, I used other MixSIAR models separately to calculate the mean diet of each age group, using an error of Residual*process (Maybe residuals only would be the more correct option?). Since individuals has just one observation per group, I tried a model with an error of process only (N=1) (according to the example of Galloway et al in the manual) with individual as a random factor (It is impossible to include fixed factors in a model with informative priors). My concern is that the resulting values for the individuals are very different compared to the model where I just included age group as random factor.

I hope this gives you an idea of how the dataset looks like and also the model that I tried to use.

Cheers,

Fernando.

[cid:image003.png@01D23104.2D22EEE0]

From: Brian Stock [mailto:notifications@github.com] Sent: 27 October 2016 16:52 To: brianstock/MixSIAR MixSIAR@noreply.github.com Cc: Fernando Chaguaceda fernando.chaguaceda@ebc.uu.se; Author author@noreply.github.com Subject: Re: [brianstock/MixSIAR] individual estimations (#89)

Hi Fernando,

First, there is no one "MixSIAR model" - there are lots of options you choose that determine the model, which depends on your data and what you're interested in.

How many individuals? How many tracers? Any fixed/random/continuous effects (besides individuals)? What error model are you using? An isospace plot of the source-consumer geometry would be helpful too.

Generally, categorical effects (like "individual") are either fixed or random, so no, there isn't a different way to estimate individual diets. If you have a continuous effect, need to have individual as a random effect, and then the "process error only (MixSIR)" option. Including individual as random effect estimates a variance term for how individuals deviate from the group mean, which means adding another residual error term in the model is inappropriate.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/brianstock/MixSIAR/issues/89#issuecomment-256664340, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AWBwBCaCiFHYOeh3JXv3xKGsQGoqGzNaks5q4LqfgaJpZM4KiMrl.

[Fernando-UppsalaUniv]

Hej,

I have two databases that I would like to analyze with different informative priors, each of them having 56 and 57 individuals, respectively. The tracers that I use are d13C and d15N, although the isotopic separation comes mainly from C (x axis in the graph). In this particular model I would just like the estimation of diet for each individual in order to correlate those values with other continuous variables. However, I used other MixSIAR models separately to calculate the mean diet of each age group, using an error of Residual*process (Maybe residuals only would be the more correct option?). Since individuals has just one observation per group, I tried a model with an error of process only (N=1) (according to the example of Galloway et al in the manual) with individual as a random factor (It is impossible to include fixed factors in a model with informative priors). My concern is that the resulting values for the individuals are very different compared to the model where I just included age group as random factor.

I hope this gives you an idea of how the dataset looks like and also the model that I tried to use.

Cheers,

Fernando.