R help - Jul 2008 - Hi~problem with the two sample test: ks2Test in the package of fbasics

Hi everyone, when I use the two sample Kolmogorov¨CSmirnov ks2Test like this:
 x=read.table("e:/x.txt")
 y=rstable(1000,alpha,beta,gamma,delta)
  
 I alway get  results as follows:
 Warning messages:
1: In ks.test(x = x, y = y, alternative = "two.sided") :
  cannot compute correct p-values with ties
2: In ks.test(x = x, y = y, exact = TRUE, alternative = "two.sided") :
  cannot compute correct p-values with ties
3: In ks.test(x = x, y = y, alternative = "less") :
  cannot compute correct p-values with ties
4: In ks.test(x = x, y = y, alternative = "greater") :
  cannot compute correct p-values with ties
 I know that the one-sample Kolmogorov¨CSmirnov test only applied to continuous
distributions, does the two-sample ks test require this? Even if it requires,
the stable distribution is continuous.
 So I also tried this: 
 x=read.table("e:/x.txt");
 y=rnorm(1000,mean,sd);
 ks2Test(x,y);
 And this:
 x=read.table("e:/x.txt");
 ks.test(x,"norm");
 Strangely,I got a total same result! I really want to know whether there are
any other sources that could  induce this result except for the continuity of
distribution? If there are, what are they?
 Any help will be much appreciated!
	[[alternative HTML version deleted]]

Hi everyone, when I use the two sample Kolmogorov¨CSmirnov ks2Test like this:
 x=read.table("e:/x.txt")
y=rstable(1000,alpha,beta,gamma,delta)

I alway get  results as follows:
Warning messages:
1: In ks.test(x = x, y = y, alternative = "two.sided") :
  cannot compute correct p-values with ties
2: In ks.test(x = x, y = y, exact = TRUE, alternative = "two.sided") :
  cannot compute correct p-values with ties
3: In ks.test(x = x, y = y, alternative = "less") :
  cannot compute correct p-values with ties
4: In ks.test(x = x, y = y, alternative = "greater") :
  cannot compute correct p-values with ties
I know that the one-sample Kolmogorov¨CSmirnov test only applied to
continuous distributions, does the two-sample ks test require this? Even if
it requires, the stable distribution is continuous.
So I also tried this:
x=read.table("e:/x.txt");
y=rnorm(1000,mean,sd);
ks2Test(x,y);
And this:
x=read.table("e:/x.txt");
ks.test(x,"norm");
Strangely,I got a total same result! I really want to know whether there are
any other sources that could  induce this result except for the continuity
of distribution? If there are, what are they?
Any help will be much appreciated!

Lily

	[[alternative HTML version deleted]]

l y <pligly <at> gmail.com> writes:
> 
> Hi everyone, when I use the two sample Kolmogorov?CSmirnov ks2Test like
this:
>  x=read.table("e:/x.txt")
> y=rstable(1000,alpha,beta,gamma,delta)
> 
> I alway get  results as follows:
> Warning messages:
 [snip]
> Strangely,I got a total same result! I really want to know whether there
are
> any other sources that could  induce this result except for the continuity
> of distribution? If there are, what are they?
> Any help will be much appreciated!
> 
> Lily
> 
  The problem is that you have ties in your *data* -- nothing
to do with the distribution you are testing against.
  You could try adding a tiny bit of random noise to your
data and see if it makes any difference in the answers -- e.g.

ks.test(rnorm(length(x),mean=x,sd=0.0001),y)

  Others may have better ideas.
  Make sure to look at ?ks.test and read the paragraph
about ties carefully ...

  Ben Bolker