Average Sample Number using Page's CUSUM procedure? - r

Question

I am trying reproduce the Table 1 results from the page 12 using the equation given the page 13.To access the journal article please click https://arxiv.org/pdf/math/0605322.pdf. The corresponding equation is given below.
My r code is give below. Am I programmed correctly?
mytest=function(n,s,c1){
t = sum(s)
k=which.max(s[19:n]>=c1)
if(k==1 && s[19]<c1)
return(c(n,0))
else
return(c(k,1))
}
for (n in c(100,200,400)){
for (i in c(-0.5, -1.0)){
a1=0
c1 = 20
asn1=0
for (m in 1:1000){
g=c(dnorm(n,0,1))
f=c(dnorm(n,i,1))
s = log(g/f)
test=mytest(n,s,c1)
a1=a1+test[2]
asn1=asn1+test[1]
}
}
out <- list(power= a1/m, asn=asn1/m)
return(out)
}
But I am getting the following errors.
Error in if (k == 1 && s[19] < c1) return(c(n, 0)) else return(c(k, 1)) :
missing value where TRUE/FALSE needed

Answer 1

The first time you call mytest, you have n=100, i=-0.5 which yields s=NaN. Therefore, you get an error on line if(k==1 && s[19]<c1) given that s[19]=NaN.
Here's a workaround, but you need to make sure it does what you expect/wish :
mytest=function(n,s,c1){
if(is.na(s)) return(c(c1,1)) # skips if NaN
t = sum(s)
k=which.max(s[19:n]>=c1)
if(k==1 && s[19]<c1)
return(c(n,0))
else
return(c(k,1))
}