New Zealand Statistical Association
2004 Conference
Submitted Talks
[ NZSA 2004 Home ] [ Abstract Submission ] [ Call for Papers ] [ Conference Program ] [ Conference Registration ] [ Conference Timetable ] [ Important Dates ] [ Invited Speaker ] [ Organising Committee ] [ Submitted Talks ] [ Travel & Accomodation ] [ VUW Campus Map ]

[ About Wellington ]

Jin Seo Cho
VUW

Testing for Identical and Independent Distributions

We generalize runs test, and provide model independent statistics testing identical and independent distribution (i.i.d.) condition. These are non-parametric statistics designed to test unspecific alternative hypotheses, as affirmed by Monte Carlo simulations. Further, we apply them to data transformations via parametric models, and show their null limiting distributions are distribution free. Thus, the same critical values can be used whether the parameter estimation errors are involved or not.

As an empirical application, we exploit our statistics and test if the Panel Study of Income Dynamics (PSID) used to estimate males' labor supply function has sample selection bias. This application exploits the fact that any data set collected by the simple random sampling procedure (SRSP) has to follow the i.i.d. condition. We show that the i.i.d. aspect of the PSID designed at the initial period (1968) has been well preserved for a long time.