acid (version 1.1)

midks.test: Kolmogorov-Smirnov Test assessing a Parametric Mixture for a Conditional Income Distribution

Description

This function performs a Kolmogorov-Smirnov test for a parametrically specified cdf composed of a mixture distribution either by cdf.mix.dag or cdf.mix.LN.

Usage

midks.test(x, y, ..., w = NULL, pmt = NULL)

Arguments

x

a vector of observed incomes.

y

a function specifying the parametric cdf.

...

arguments to be passed to y.

w

the weights of the observations contained in y.

pmt

point mass threshold equivalent to thres0 in y.

Value

• statistic
  returns the test statistic.
• method
  returns the methodology - currently always One-sample KS-test.
• diffpm
  the difference of the probability for the point mass.
• diff1
  the upper difference between for the continuous part of the cdfs.
• diff2
  the lower difference between for the continuous part of the cdfs.

References

Sohn, A., Klein, N. and Kneib. T. (2014): A New Semiparametric Approach to Analysing Conditional Income Distributions, in: SOEPpapers, No. 676.

Examples

# parameter values
pi0.s<-0.2
pi1.s<-0.1
thres0.s<-0
thres1.s<-25000
mu.s<-20000
sigma.s<-5
nu.s<-0.5
tau.s<-1

# generate sample
n<-100
s<-as.data.frame(matrix(NA,n,3))
names(s)<-c("cat","y","w")
s[,1]<-sample(1:3,n,replace=TRUE,prob=c(pi0.s,pi1.s,1-pi0.s-pi1.s))
s[,3]<-rep(1,n)
for(i in 1:n){
  if(s$cat[i]==1){s$y[i]<-0
  }else if(s$cat[i]==2){s$y[i]<-runif(1,thres0.s,thres1.s)
  }else s$y[i]<-rGB2(1,mu=mu.s,sigma=sigma.s,nu=nu.s,tau=tau.s)+thres1.s
}

# midks.test
midks.test(s$y,cdf.mix.dag,pi0=pi0.s,thres0=thres0.s,pi1=pi1.s,thres1=thres1.s,mu=mu.s,
sigma=sigma.s,nu=nu.s,tau=tau.s,w=s$w,pmt=thres0.s)$statistic