An anova method for the "fit.bd" object class (and how to write your own)

Over the past couple of days I've either shown or used new S3 generic anova methods for different phytools object classes.

These are relatively straightforward to write, and since they can make model comparison in phytools (sometimes) much easier, I thought I'd illustrate an example function.

Here is an anova method for the object class "fit.bd". ("fit.bd" objects are created by the phytools functions fit.bd, for fitting a birth-death model with incomplete taxon sampling to a reconstructed phylogeny; and fit.yule.)

I'll add this to phytools shortly.

anova.fit.bd<-function(object,...){
    fits<-list(...)
    nm<-c(
        deparse(substitute(object)),
        if(length(fits)>0) 
            sapply(substitute(list(...))[-1],deparse)
    )
    fits<-c(list(object),fits)
    logL<-sapply(fits,logLik)
    df<-sapply(fits,function(x) attr(logLik(x),"df"))
    AIC<-sapply(fits,AIC)
    weight<-unclass(aic.w(AIC))
    result<-data.frame(logL,df,AIC,weight)
    rownames(result)<-nm
    colnames(result)<-c("log(L)","d.f.","AIC","weight")
    print(result)
    invisible(result)
}

Now let's use it with a simulated case in which the true death (extinction) rate is exactly half the birth (speciation) rate.

tree<-pbtree(b=0.1,d=0.05,t=100)
tree<-drop.tip(tree,getExtinct(tree))
plot(ltt(tree,plot=FALSE),bty="n",log.lineages=FALSE,
    log="y",cex.axis=0.8,las=1)
grid()

Fit our models.

birth_death<-fit.bd(tree)

## Warning in nlminb(c(init.b, init.d), lik.bd, t = T, rho = rho, lower = rep(0, : NA/NaN function evaluation

birth_death

## 
## Fitted birth-death model:
## 
## ML(b/lambda) = 0.08452 
## ML(d/mu) = 0.04266 
## log(L) = 22.6402 
## 
## Assumed sampling fraction (rho) = 1 
## 
## R thinks it has converged.

yule<-fit.yule(tree)
yule

## 
## Fitted Yule model:
## 
## ML(b/lambda) = 0.06173 
## log(L) = 20.0631 
## 
## Assumed sampling fraction (rho) = 1 
## 
## R thinks it has converged.

Compare using new anova method.

anova(yule,birth_death)

##               log(L) d.f.       AIC    weight
## yule        20.06310    1 -38.12619 0.1712015
## birth_death 22.64023    2 -41.28046 0.8287985