Order statistics has an important role in statistical inference. The m
ain purpose of this paper is to investigate order statistics, and also
explore its applications in the analysis of nonstationary time series
. Our results show that linear functions of order statistics for a lar
ge class of time series are asymptotically normal. The methods of proo
f involve approximations of serially dependent random variables by ind
ependent ones. The problems of testing for the existence of a linear t
rend and the problem of testing randomness versus serial dependence ar
e considered as applications.