Über die Anzahl von Zufallspunkten mit typ-gleichem nächsten Nachbarn und einen multivariaten Zwei-Stichproben-Test (Q800661)

For independent s-variate samples \(X_ 1,...,X_ m\) i.i.d. f(.), \(Y_ 1,...,Y_ n\) i.i.d. g(.), where the densities f(.), g(.) are assumed to be continuous on their respective sets of positivity, consider the number \(T_{m,n}\) of points Z of the pooled sample (which are either of 'type X' or of 'type Y') such that the nearest neighbor of Z is of the same type as Z. We show that, as m,n\(\to \infty\), \(m/(m+n)\to \tau\), \(0<\tau <1\), f(.)\(\equiv g(.)\), \(\lim Var T_{m,n}/\sqrt{m+n}=\sigma^ 2(\tau,s),\) independently of f(.). An omnibus test for the two sample problem 'f(.)\(\equiv g(.)\) or f(.)\(\not\equiv g(.)?'\) may be obtained by rejecting the hypothesis f(.)\(\equiv g(.)\) for large values of \(T_{m,n}\).