Professional Documents
Culture Documents
Data
Short Resume
Name- Aakriti Pandey
Qualification Details:
B.Sc. Statistics Hons. from Banaras Hindu University with 85.5% (2010-2013)
In life testing experiments, censoring is a common feature and may occur naturally
or owing to some constraints.
The most common censoring schemes used in practice are Type-I and Type-II.
Mixture of Type-I, Type-II led to Type-I hybrid censoring scheme and Type-II hybrid
censoring scheme having time of the termination T ∗ = min{Xm:m:n , T } and
T ∗ = max (Xm:m:n , T ).
The Type-I hybrid censoring scheme keeps the termination time of the experiment
below a prefixed value by forfeiting the efficiency, whereas Type-II hybrid censoring
ensures efficiency more than the prefixed level but forfeits the termination time.
Therefore, the need for a censoring scheme controlling termination time and
efficiency was felt simultaneously. [Cho et al.(2015)Cho, Sun, and Lee] introduced
the generalized progressive hybrid (GPH) censoring scheme which terminates at
max(Xk ,min(Xm ,T )).
Experiment Start
Case I: X(m) > X(k) > T
∗
RD
R1 R2 Rk RD
Number of Removals
Experiment Start
Case II: X(m) > T > X(k)
Rk Rm
R1 R2
Number of Removals
Experiment Start
Case III: T > X(m) > X(k)
(
1, if J = k, m
where, W (α, β) = ∗
RD+1
[1 − F (T )] , if J = D.
We have considered different removal pattern such as
Sm:n (1) : All the removals are at the last failure, i.e., Rm = n − m.
Sm:n (2) : All the removals are at the first failure, i.e., R1 = n − m.
Sm:n (3) : The removals are at the first and last failure, i.e., R1 = Rm = (n − m)/2.
Sm:n (4) : The removals are at middle failure, i.e., Rm/2 = Rm/2+1 = (n − m)/2.
density
Y[8]= (155.1725, 426.6925) Y[6]= (187.0323, 311.3068)
0.010 0.02
Y[10]= (194.9043, 544.8581) Y[8]= (195.4682, 356.7346)
Y[12]= (236.0470, 696.0977) Y[10]= (217.3346, 433.2698)
Y[14]= (284.1179, 881.8822)
0.000 0.00
Figure 2.1: Density plot of one sample predicted order statistics with their respective 95%
confidence interval.
0.05 0.04
Two sample future prediction Two sample future prediction
for Dataset−1.2 for Dataset−2.3
0.04
0.03
Order 95% CI Order 95% CI
Y[2]= (0.0885, 31.8225) Y[2]= (0.1328, 40.0638)
0.03
Y[4]= (0.5265, 50.5970)
density
density
Y[6]= (0.8087, 64.3023) 0.02 Y[6]= (4.6168, 77.3147)
Y[8]= (4.6441, 77.0188) Y[8]= (11.9088, 96.7540)
0.02
Y[10]= (11.8215, 95.9900) Y[10]= (15.8471, 110.7657)
Y[20]= (27.0634, 159.4856) Y[20]= (46.9672, 197.8473)
0.01
0.01
0.00 0.00
Figure 2.2: Density plot of two sample predicted order statistics with their respective 95%
confidence interval.
There may be several causes which leads to the failure of an item. These causes
tend to compete with each other. Hence, in the statistical literature this is popularly
known as competing risk problem.
The objective of a competing risk problem is to find the lifetime distribution of a
particular cause, in presence of the other causes.
The assumption of prefixed removal is not realistic for real life phenomenon. In a
clinical trial, the number of patients dropping out is beyond the control of
experimenter and cannot be predetermined.
No one has attempted the inferencial procedure for GPH censored competing risk
data considering removals to be random. This motivated us to proceed with the
inferencial procedure for GPH censored competing risk data wherein removals are
governed by beta binomial probability law.
The MPS procedure provides more precise estimates than those obtained from
maximum likelihood and bootstrap procedures.
The Bayesian procedure delivers more accurate and precise estimates of the
parameters even if we consider the vague prior.
The width of the HPD interval is smaller than asymptotic and bootstrap confidence
intervals. So, with this, we propose to use HPD interval under considered problem.
For the estimation of parameter of the distribution, one can use removal pattern
Sm:n (1) in place of Sm:n (2) and Sm:n (4) in place of Sm:n (3) and for estimation of the
acceleration factor, one can prefer removal pattern Sm:n (2) to Sm:n (1) and Sm:n (4) to
Sm:n (3) .
Since, for the large values of k, the MSEs of estimators of ϵ and ζ decrease.
Therefore, it is suggested to keep the value of k high. Similarly, we suggest to keep
the value of T small.
For large values of ζ, the ETT is noted to be smaller however, as τ increases the
ETT increases, keeping other factors fixed. Thus, a reduction in the ETT is
expected for large values of ζ and small values of τ .
In competing risk data, we observed that the effect of increasing n, m, k, and T
decreases MSEs and average confidence length of estimators.
In ALT, the acceleration is performed at one step but, the work can be extended by
applying acceleration at multiple steps. Also, It would be salutary to find the
optimum time of the acceleration.
Statistical analysis of competing risks with masked failure causes based on GPH
censoring with random removals.
Different method of estimation for a model under SSPALT based on GPH censoring.
Statistical Analysis for Generalized Progressive Hybrid Censored Data from Lindley
Distribution under Step-Stress Partially Accelerated Life Test Model.
Austrian Journal of Statistics 50.1 (2021): 105-120.