Simulation of sequential Cp criteria for the regression
                     model selection

           Nataliya A. Tsepkova, Alexander V.Tsukanov

      The paper is devoted to the Monte-Carlo research of
the  sequential  Cp  criteria for  the  regression  model
selection.  The  selection  procedure is  connected  with
testing statistical   hypothesis with given level of  the
first and second type errors.
      The  case of  the  nested  set of regression models
and  the  normally distributed error with  zero  mean  is
considered.  The  model is selected on the  base  of  the
small  sample. The quality of the selection procedure  is
estimated by  the  probabilities of  error.
      The  properties  of  the   sequential   Cp criteria
with   varying  parameter Co for the model selection have
been  studied. Parameter Co is the function of the  first
and  second  type errors.  It  is  proposed  to  continue
the  experiment  until the value of Cp criteria  for  any
model  larger  than  Co.
     Simulation results of  two procedures were compared.
The first one is the sequential selection  procedure with
the  same level of the errors as for the second one-stage
procedure. The results of Monte-Carlo study   shows   the
possibility  to reduce the mean number of observations by
sequential criteria.