Plotting partial residuals from a glmmADMB model

Dear Dr. Bolker,
 Thank you very much for your response. After trying to implement your suggestion, I'm unfortunately stuck. It appears that the getME function does not work with glmmADMB. I get the following error message:

Error in UseMethod("getME") : 
  no applicable method for 'getME' applied to an object of class "glmmadmb"

A potential work around to this may be to use the "predict" function, which can generate the components, but I'm not sure if this is equivalent. Also, I'm having trouble following the steps that you outlined in your email to generate the partial residuals. Would you be willing to work through how to generate the partial residuals for the fixed effect "FoodTreatment" in the model below that uses the Owl data set?

 ##Using the Owl Data
om <- glmmadmb(SiblingNegotiation~FoodTreatment+ArrivalTime+SexParent+ (1|Nest),family="nbinom1",data=Owls)

Thank you,
Frances

-----Original Message-----
From: Ben Bolker [mailto:bbolker at gmail.com] 
Sent: Thursday, December 14, 2017 9:54 PM
To: Sivakoff, Frances <sivakoff.3 at osu.edu>
Cc: r-sig-mixed-models at r-project.org
Subject: Re: [R-sig-ME] Plotting partial residuals from a glmmADMB model

You'll have to find a way to make "partial predictions". I don't think there's anything built in for this.  *Very* briefly, considering only fixed effects, if you retrieve the X matrix
(getME(fitted_model,"X")) and the fixed-effect parameters (fixef(fitted_model)), you can drop any columns/parameters you want and do exp(X %*% beta) with the remaining columns/parameters to get a prediction that includes some but not all of the predictors. Subtracting the observed value should get you the partial residuals ...

I would like to use ggplot2 to plot the partial residuals of an indicator (0 or 1) independent variable in a generalized linear mixed model fit with a "nbinom1" family using glmmadmb. My model has a response variable that is a count, 3 explanatory variables that are continuous, and 7 indicator variables that are 0 when a particular heavy metal is not detected and 1 when it is detected above a threshold value. I'd like to plot the partial residuals of the various independent variables. I think that the model below using the Owl data would be a good example data set for how to do this. The model below has a response variable that is a count, an explanatory variables that is continuous (arrivalTime), two categorical variables (FoodTreatment and SexParent), a random effect of Nest, and uses a "nbinom1" family.

##Using the Owl Data
om <- glmmadmb(SiblingNegotiation~FoodTreatment+ArrivalTime+SexParent+
                 (1|Nest),family="nbinom1",data=Owls)

Could you please suggest a method for plotting the partial residuals of the explanatory variables.

Thank you,
Frances


        [[alternative HTML version deleted]]

_______________________________________________
R-sig-mixed-models at r-project.org mailing list 
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models