n=c(40,50,40) x=c(1,2,3) y=c(11,32,33) rin=matrix(c(y,n-y),3,2) res=glm(rin~x, family=binomial(link="probit")) summary(res) #Median lethal dose LD=-res$coef[1]/res$coef[2] LD #House sparrows dispersal ~ wing length #Make design #Make one specific dataset set.seed(555) N=1000; #Set true parameters beta0 = 2; beta1 = -0.5; #Sample some covariates/measurments x1=rnorm(N,5,0.3) #Calculate mu eta = beta0 + beta1*x1; #Use probit link p=pnorm(eta,0,1); #Simulate data y=rbinom(rep(1,N),rep(1,N),p) sum(y) #R wants indata as [y, n-y] respin=matrix(c(y,rep(1,N)-y),N,2) resultprobit = glm(respin~x1, family=binomial(link="probit")) summary(resultprobit) fittedprobit=resultprobit$fitted; plot(x1,fittedprobit) #Use logit link instead resultlogit = glm(respin~x1, family=binomial(link="logit")) summary(resultlogit) fittedlogit=resultlogit$fitted; points(x1,fittedlogit,col='blue') #Use cloglog link instead resultcloglog = glm(respin~x1, family=binomial(link="cloglog")) summary(resultcloglog) fittedcloglog=resultcloglog$fitted; points(x1,fittedcloglog,col='green')