Abstract. This paper deals with the problem of reconstructing missing data in a left type-II censoring scheme, where the underlying distribution is the Weibull distribution. Frequentist and Bayesian approaches are adopted in order to provide some point reconstructors for the past failure times. The problem of determining reconstruction intervals for the past failure times is also considered. The investigation includes an example of application to real data and various comparisons based on Monte Carlo simulations.

w l}Uqm OQwmQ x=Qty Q : }eDt |=yQw=DWo CLD u}ov=}t |QDt=Q=B=v |=yu=Qm uQDWvoQwt p@t=o |rQ=i pOt |OtL= QiaH w |v}t= / | DQt OyWt |UwOQi x=oWv=O |=yxQ=t; R= |=xr=@vO |UQQ@ \ki xm|v=tR 'xOWCiH |iO=YD |=yQ : }eDt R= xr=@vO l} QO "xO}mJ (concomitant) x=Qty Q : }eDt u=wva CLD xQ=t; Qy =@ Q_=vDt swO |xir-wt 'OW=@ Q_v Ot xirwt wO R= |m} =yQwUv=U '=yOQwmQ '|@}DQD |=yxQ=t; u=wD|t =yxQ=t; u}= |xrtH R= "OQ}o|t Q=Qk |UQQ@ OQwt xQ=t; u; QO "CU= =yOQwmQ x=Qty |=yQ : }eDt |xr=@vO X=wN |UQQ@ hOy xr=kt u}= QO "OQ@ s=v =Q xQ}e w QDlJwm =} w QDnQR@ |r@k C=Oy=Wt R= xm OvDUy |Q}O=kt =yOQwmQ '|iO=YD |=yQ : }eDt R= xr=@vO l} |v}r=@ |=yx}=tR; =} w |WRQw C=k@=Ut x@ \w@ Qt |=yxO=O '|a}@] |=yxO}OB R= |Q=}U@ "OvDUy CiH R= |=xr=@vO |=Q@ Qw_vt u}O@ "OvDUy |UQQ@ p@=k 'p=tDL= |=ywor= u}vJ u}= ?r=k QO |iO=YD |=yQ : }eDt |=yQw=DWo X=wN (X; Y) |iO=YD |=yQ : }eDt =@`} RwDsy w pkDUt C=Oy=Wt uQDWvoQwt p@t=o |rQ=i pOt CLD w |rm Cr=L QO X |iO=YD Q : }eDt |=yOQwmQ =@ Q_=vDt Y |QDt=Q=B=v |=yu=Qm xt=O= QO "CU= xDiQo Q=Qk |UQQ@ OQwt Farlie-Gumbel-Morgenstern "Ov=xOW xU}=kt sy =@ w xOt; CUO x@ Cw=iDt |=yxw}Wx@ w q=@ pOt CLD u}ov=}t |Q @=Q @= v &uQ D W v oQw t p @ t= o | rQ= i pO t &Qw= D W o &x=Q t y Q : } e D t &q= @ OQw mQ "|O } r m u= oS=w "|QDt=Q=B=v u=Qm &TDQ=wW |Wwm "C=@D=mt Q=OxOya |xOvU} wv