Page 110 - Contributed Paper Session (CPS) - Volume 1
P. 110

CPS1167  Indrani B.
                  resulting prediction. ̂ is given by (13) with the denominator modified to .
                  Then, the linearized UMVUE ̂  corresponding to ̂ is obtained by solving (10)
                                               1
                  when  is substituted by this ̂ .

                                                                
                  Case  2  : ( =  and  = 1, … ,  1)   The  CMP   :  of  :   is  given  by  (18)  in
                              1
                  which the linearized UMVUE ̂ given by (16) with the denominator modified
                  to  and the linearized UMVUE ̂  is obtained by solving (15) when  in there
                                                  1
                  is substituted by this ̂.

                                                                            
                  Case 3 : (0 ≤  ≤  − 1 and  =  + 1, … , ) The CMP   :  of  :   is given
                                 1
                                                    1
                  by
                                                       − ̂ 2
                                                      
                                      = ̂ + ̂ log [   ̂ +   ]
                                           2
                                    :                   , 
                                                                                         (19)

                  3.  Results
                      Progressive  Type-II  censored  data  under  the  step-stress  setting  are
                  generated and the values of MLP and CMP of    are computed under Khamis
                                                               
                  and  Higgins  Model.  A  numerical  study  is  carried  out  to  compare  the
                  performances of these MLP and CMP in terms of their Mean Square Prediction
                  Errors (MSPEs) and Prediction Intervals (PIs). Derivations of the MSPEs and PIs
                  of the MLP and CMP are complicated and so we used simulations to get those.
                  Using simulation studies, standard errors of these    were generated and the
                                                                    
                  PIs were constructed for each of the predictors MLP and CMP.

                  4.  Discussion and Conclusion
                      In this article, we have derived the MLP and CMP for the survival times of units
                  from the Weibull distribution which are progressively Type II censored under the
                  Khamis and Higgins model for the simple step-stress data. Simulation studies are
                  used to illustrate and compare the methods developed in this article. Simulation
                  studies show that the predicted values for CMP are generally closer to their actual
                  values than the corresponding predicted values for MLP. This is particularly true
                  for larger sample size, larger number of uncensored observations and for delayed
                  censoring scheme as long as the number of predicted observations are not very
                  small.

                  References
                  1.  Banerjee, A. and Kundu, D. (2008). Inference Based on Type-II Hybrid
                      Censored Data from a Weibull Distribution, IEEE Transactions on
                      Reliability 57, pp. 369-378.
                  2.  Khamis, I. H. and Higgins, S. H. (1998). A new model for step-stress
                      testing, IEEE Transactions on Reliability 47, pp. 131-134.

                                                                    99 | I S I S   W S C   2 0 1 9
   105   106   107   108   109   110   111   112   113   114   115