}x$corr=apply(M, 2, mean)return(invisible(x))### Corrélation du modèle glm (Boyce et al. 2002) ###corr.RSF=function(x, presence=ncol(x$df), k=5, ...) {par(mfrow=rev(n2mfrow(k)))kdf=kpart(x$df, k)f1=function(i) {hsp=predict(x, newdata=kdf[[i]]$test[,-presence])$HSquant=quantile(hsp[kdf[[i]]$test[,presence]==1],probs=c(0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1))quant[1]=0quant[11]=1G=hist(hsp, br=quant, plot=F)H=hist(hsp[kdf[[i]]$test[,presence]==1], br=quant, plot=F)ratio=(H$counts/sum(H$counts))/(G$counts/sum(G$counts))plot((1:10)/10,ratio, xlab='HS', ylab='Utilisation relative',type='h')A=cor.test(1:10, ratio, method='spearman')B=cor.test(1:10, ratio, method='kendall')C=cor.test(1:10, ratio)abline(h=1)return(list(A$est, A$p.value, B$est, B$p.value, C$est,C$p.value))}M=matrix(unlist(lapply(1:k,f1)), nrow=k, byrow=T,dimnames=list(1:k,c('S.est.cor','S.p.value','K.est.cor','K.p.value','P.est.cor','P.p.value')))par(mfrow=c(1,1))x$corr=apply(M, 2, mean)return(invisible(x))}### Générique performance ###perf=function(x, ...)UseMethod("perf")### Performance de l'ENFA ###perf.ENFA=function(x, percent=scan(nmax=1), presence=ncol(x$df), k=5, ...){x=MPA(x, percent)x=corr(x, presence, k)return(invisible(x))}### Performance du glm ###perf.RSF=function(x, percent=scan(nmax=1), presence=ncol(x$df), k=5, ...) {x=MPA(x, percent)x=corr(x, presence, k)return(invisible(x))}### roc-plot ###roc.plot=function(y, ypred, spa, add=FALSE, ...){seuil=seq(0,1,spa)roctable=matrix(0,ncol=3,nrow=length(seuil))for(i in 1:length(seuil)){a=matrix(0,ncol=2, nrow=2)37
}a[1,1]=length(ypred[(ypred=seuil[i])&(y==1)])a[1,2]=length(ypred[(ypred>seuil[i])&(y==0)])a[2,1]=length(ypred[(ypred
- Page 1 and 2: Basille, M. 2004. Le lynx, l'ENFA e
- Page 4: RemerciementsJe voudrais remercier
- Page 7 and 8: premier ordre définit la répartit
- Page 9 and 10: 2. Matériel et méthodes2.1. Donn
- Page 11 and 12: lynx juvéniles (Vandel, 2001). Sep
- Page 13 and 14: 2.2. Méthodes d'analyse2.2.1. L’
- Page 15 and 16: a)Ub)Si s est supérieur à la méd
- Page 17 and 18: GLM-ENFA. En s'appuyant sur un plan
- Page 19 and 20: 2.3.2. Choix du meilleur indiceLa m
- Page 21 and 22: 3. Résultats3.1. Performance des m
- Page 23 and 24: Le test de Monte-Carlo effectué su
- Page 25 and 26: toutes les valeurs d'habitat supér
- Page 27 and 28: pForets pIFN pHydro pNatpOuver pRoa
- Page 29 and 30: 4. DiscussionCe travail s'inscrit d
- Page 31 and 32: Additive Model). Ce résultat montr
- Page 33 and 34: 5. BibliographieBoyce, M.S. & McDon
- Page 35 and 36: Reutter, B.A., Helfer, V., Hirzel,
- Page 38 and 39: }liste=list(df=df, names=names, pr=
- Page 40 and 41: ### Biplot de l'ENFA ###biplot.ENFA