ALL Metrics
-
Views
-
Downloads
Get PDF
Get XML
Cite
Export
Track
Research Article

Computational Evaluation For The Survival Times Of Patients Under Truncated Life Tests: Simulation And Application

[version 1; peer review: 3 approved with reservations]
PUBLISHED 27 Jun 2024
Author details Author details
OPEN PEER REVIEW
REVIEWER STATUS

Abstract

Background

Statistical process control (SPC) is utiled in manufacturing and service applications. It is used to monitor the behavior of the process, also to discover and solve the issues in internal systems. SPC allows immediate corrective actions to prevent defects and maintain process stability. There are two types of control charts are available in the literature: variable control charts and attribute control charts.

Methods

The use of attribute control charts for two positively skewed distributions such as power function distribution and weighted power function distribution is discussed in this research paper. We compared the behavior of both types of control charts and presented the results in form of tables.

Results

We conclude that control charts under a weighted power function distribution perform better than control charts using a power function distribution. Furthermore, these results will help in the early detection of errors that enhance the process reliability of the telecommunications and financing industries.

Keywords

power function distribution, weighted power function distribution, attribute control chart, truncated life test, average run length, simulation

1. Introduction

Statistical process control is used as an effective tool to monitor the manufacturing process and prevent products from being out of the given provisions. Two different types of control charts are available in the literature: variable control charts and attribute control charts. When numerical measurements of the variable of interest from any process are available, the use of a variable control chart is preferred. However, there are many processes that have a variable of interest in the form of attributes and are not measurable. In such situations, it is preferable to use attribute control charts instead of variable control charts.

Attribute control charts are used to monitor a process with two possible outcomes: conforming and nonconforming. The basic condition for using these charts was the assumption of normality. However, in real-life situations, whenever a process is working, the outcome may not follow a normal distribution.

It is possible that the shape of the distribution is skewed. For such cases, many studies, such as Chang and Bai,1 Lin and Chou,2 and Rao and Subbaiah,3 have been conducted. Work on attribute control charts considering different skewed distributions has also been done by Aslam et al.,4 Karagöz,5 and Noiplab and Mayureesawan.6 We can also refer to Zaka et al.,7 who modified the attribute control chart for the shape of the process using the power function and survival-weighted power function distribution.

The shape of any distribution can be studied by estimating its shape parameter. Several estimators have been developed to estimate the shape parameters of any distribution. Zaka et al.7 introduced and studied an attribute control chart for the shape of the process using the power function and survival-weighted power function distribution. Dallas8 introduced the power function distribution (PFD) as a powerful tool to study many processes in reliability and health sciences, which was further assumed and modified to increase its applicability in unequal probability theory by Zaka et al.9 as the weighted power function distribution (WPFD). Jabeen and Zaka10 proposed modified control charts based on shape parameters.

We introduce the PFD and WPFD in Section 2. In Section 3, we propose an attribute control chart for PFD and WPFD. A discussion of the use of the proposed control chart is presented in Section 4.

2. Methods

2.1. Distribution function for PFD and WPFD

The weighted power function distribution (WPFD) was used by Zaka et al9 by modifying the existing Power function distribution (PFD). The cumulative distribution function and probability density fuction for PFDand WPFD is given asintroduced

(1)
F(x1)=(x1β)γandf(x1)=γx1γ1βγ,0<x1<β,andβ,γ>0.
(2)
F(x2)=(x2β)2γandf(x2)=2γx22γ1β2γ,0<x2<β,andβ,γ>0.
where βandγ are the scale and shape parameters.

The mean of PFD and WPFD are respectively

μ1=βγγ+1andμ2=2βγ2γ+1

2.2. Average Run Length ARL for proposed control monitoring using attribute control chart for PFD and WPFD

The control limits for attribute control charts to monitor the shape parameter of weighted Power fuction distribution is illustrated in from Zaka et al.7 and similarly the control limits for attribute control charts is given by Aslam et al.11 , the proposed control limits are given as:

UCL=np0+Lnp0(1np0)
LCL=max[0,np0Lnp0(1np0)].

where p0 is the failure probability, n is the sample size, and L is the control coefficient. Thus, it is obtained from (1) and (2) as follows:

Po1=(to1βo1)γ01,Po1=(aμo1βo1)γ01;whereμo1=βo1γo1γo1+1
Po2=(to2βo2)2γ02,Po2=(aμo2βo2)2γ02;whereμo2=2βo2γo22γo2+1
where βo1,βo2 are scale parameters and γ01,γ02 are the shape parameters for the two proposed distributions.

The control limits are presented as

UCL=D¯+LD¯(1D¯n)
LCL=max[0,D¯LD¯(1D¯n)]
where D¯ is the average number of failures over the subgroups.

We can find the ARL using the following formulae

ARLoj=11pinoj
where the probability of the control chart for being in the state of control is
pinoj=P(LCLDUCL|Poj)=d=LCL+1UCL(nd)Pojd(1Poj)nd

2.1.1. The probabilities of failure for PDF in out of control process are

→ When the shape parameter varies → P11=(aβo1γ11γ11+1βo1)γ11

→ When the scale parameter varies → P21=(aβ11γo1γo1+1β11)γo1

→ All the parameters vary → P31=(aβ11γ11γ11+1β11)γ11

2.1.2. The probabilities of failure for WPFD in out of control process are

→ When the shape parameter varies → P12=(a2βo2γ122γ12+1βo2)2γ12

→ When the scale parameter varies → P22=(a2β12γ022γ02+1β12)2γ02

→ All the parameters vary → P32=(a2β12γ122γ12+1β12)2γ12

We used γ11,γ12 when the shape parameters varied fromγo1,γ02 for the first and second distributions, respectively. We also used β11,β12 when the shape parameters varied from βo1,βo2 for the first and second distributions, respectively.

2.3. Case 1 (when we assume shift in scale parameter)

When the parameter is shifted from β1=Cβ0, the probability in the control process is

pin1j=P(LCLDUCL|P1j)=d=LCL+1UCL(nd)P1jd(1P1j)nd

The ARL for the shifted process is given as

ARL1j=11pin1j

2.4. Case 2 (when we assume a shift in the shape of a parameter)

We adopt γ1=Cγ0. The probability of in-control for the shifted process is

pin2j=P(LCLDUCL|P2j)=d=LCL+1UCL(nd)P2jd(1P2j)nd

The ARL to compare the efficiency of the process after the shift is

ARL2j=11pin2j

2.5. Case 3 (when we assume a shift in the scale of a parameter)

When the scale and shape parameters are shifted as β1=Cβ0 and γ1=Cγ0we get the probability

pin3j=P(LCLDUCL|P3j)=d=LCL+1UCL(nd)P3jd(1P3j)nd

The ARL for the shifted process is

ARL3j=11pin3j

Zaka et al.7 provided the simulation steps to conduct the monitoring of the shape parameter of weighted power function distribution. By considering these steps we assumed different observations of the ARL, shape parameter, and a. For these observations of ARL, we took the values of the shape and scale parameters of the control chart and sample size (n) in such a way that the ARL from equation (ARL0) was close to R0. Using equation (ARL1) and the values fixed in equation (2), we found ARL1 using different values of c,γ0, R0, a, and n, which are in Tables 1-4.

Table 1. Average run length for the attribute control chart setting R0=300 and sample size=5.

For n=5PFDWPFDPFDWPFD
γo1=0.5γo2=0.5γo1=1γo2=1
Shift in parameterR0=300 d=0.306388
r=2.843
R0=300 d=0.143885
r=3.977
R0=300 d=0.333058
r=3.627
R0=300 d=0.20458
r=3.951
1.0300.002300.004300.029300.016
0.95244.815220.501233.793216.646
0.9199.177161.995181.846156.447
0.85161.511118.963141.156112.988
0.8130.49587.326109.32581.619
0.75105.01564.07984.46158.981
0.784.13747.00565.07142.646
0.653.18925.27738.24822.354
0.532.80013.59522.12411.789
0.419.5837.33112.5336.287
0.311.1883.9856.9063.423
0.25.9922.2143.6661.942
0.12.8771.3121.8581.215

Table 2. Average run length for the attribute control chart setting R0=300 and sample size=10.

For n=10PFDWPFDPFDWPFD
Shift in parameterγo1=0.5γo2=0.5γo1=1γo2=1
R0=300 d=0.18446
r=3.063
R0=300 d=0.13733
r=3.108
R0=300 d=0.2342
r=2.808
R0=300 d=0.1408
r=5.408
1.0300.012300.005300.123300.025
0.95216.107201.670207.981201.823
0.9155.594135.888144.323135.926
0.85111.98091.820100.31091.691
0.880.56762.25269.85261.984
0.7557.95642.37648.75342.019
0.741.69028.98534.12028.589
0.621.59613.81516.89013.444
0.511.2376.8168.5316.530
0.45.9103.5434.4473.349
0.33.1832.0012.4431.882
0.21.8081.2921.4761.234
0.11.1641.0291.0681.017

3. Results and Discussion

To study the proposed control charts (Section 3), we assume the shape parameter as γo1=0.5andγo2=1. We also assumed a target ARL value of 300 and simultaneously for the in-control process. We assume a shift of 0.1 in the shape parameter for PFD and WPFD, and n = 5, 10, and 20. The results are presented in Tables 1-3 and Figures 1-2. We observed a rapid decrease in ARL for WPFD compared to PFD, with a decrease in C and an increase in n. We see that for γo1=0.5 to γo1=1, there is an early detection in out of control ARLs for both the PFD and WPFD control chart. We also observed that with an increase in the size of the sample and the value of the shape parameter, the WPFD proved to be more reliable for use instead of PFD.

Table 3. Average run length for the attribute control chart setting R0=300 and sample size=20.

PFDWPFDPFDWPFD
Shift in parameterγo1=0.5γo2=0.5γo1=1γo2=1
R0=300 d=0.12797
r=2.972
R0=300 d=0.155871
r=3.011
R0=300 d=0.21199
r=3.41
R0=300 d=0.26795389
r=3.4
1.0300.028300.003300.003300.000
0.95185.935179.961183.113182.826
0.9116.035108.906112.665112.212
0.8572.99466.57969.95869.450
0.846.34441.18543.89943.411
0.7529.73825.82727.88327.454
0.719.32016.45517.96017.604
0.68.5377.0867.8477.626
0.54.0883.3833.7523.627
0.42.1931.8622.0341.968
0.31.3801.2411.3111.281
0.21.0681.0311.0481.040
0.11.0021.0001.0011.000
ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure1.gif

Figure 1. ARLs for attribute control charts of PFD and WPFD for γ=0.5 and K=300.

ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure2.gif

Figure 2. ARLs for attribute control charts of PFD and WPFD under γ=1 and R0=300.

3.1. Simulated data

In this section, we generated data from the PFD and WPFD. Let n=10 and R0=300,γoi=0.5and βoi=1, where i=1, 2. We generated the first 10 observations from the process under control and the next 20 observations from the shifted process using shift = 0.85, as presented in Table 4. We used the life test termination times t01=0.31259*μo1and t02=0.2235101*μo2. Figures 3 and 4 show the 30th observation (10th observation after the shift) using the WPFD control chart. We conclude that WPFD is more efficient than the PFD control chart.

Table 4. Simulated data from PFD and WPFD.

PFDWPFD
63
22
45
54
26
13
34
42
55
75
36
22
43
52
41
23
15
45
55
62
73
64
36
57
41
45
64
32
25
54
78
32
55
45
42
33
25
11
53
42
ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure3.gif

Figure 3. Simulation for PFD.

ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure4.gif

Figure 4. Simulation for WPFD.

3.2. Real Life data

We used survival time data for a group of patients given by Bekker et al.12 The data consisting of survival times (in years) for 46 patients were: .0470, .1150, .1210, .1320, .1640, .1970, .2030, .2600, .2820, .2960, .3340, .3950, .4580, .4660, .5010, .5070, .5290, .5340, .5400, .6410, .6440, .6960, .8410, .8630, 1.0990, 1.2190, 1.2710, 1.3260, 1.4470, 1.4850, 1.5530, 1.5810, 1.5890, 2.1780, 2.3430, 2.4160, 2.4440, 2.8250, 2.8300, 3.5780, 3.6580, 3.7430, 3.9780, 4.0030, 4.0330.

We used the maximum likelihood estimator (MLE) to estimate the PFD and WPFD parameters. The statistics used to show the application of the control charts are shown in Figures 5 and 6. Zaka et al.9 showed that the above data follow the WPFD with the shape parameterγo1=0.3087394. We also assumed that ao1=0.35548. We obtained the control coefficient of the proposed chart as ko1=2.978, n=15, and p01 = 0.465220. From Figure 5, we observe that the control chart based on the PFD shows an in-control situation. All the points lie between the control limits.

ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure5.gif

Figure 5. Real life analysis for PFD.

ef1e2c4d-7d36-46db-9d89-e6c64831e86d_figure6.gif

Figure 6. Real life analysis for WPFD.

From Figure 6, the WPFD control chart is presented with shape parameter γo2=0.3087394. The plotting statistics are shown in Figure 6, which are used to apply the proposed control chart. The data are known to follow the WPFD with shape parameterγo2=0.308739,ao2=0.262064 and control coefficients ko2=2.869, n=15, and p02 = 0.423427. From Figure 6, we see that all points are in the control area, but the 30th observation shows an out-of-control situation.

We see that it is better to use the WPFD-based control chart to control the process than the PFD control chart.

4. Conclusion

In this paper, we proposed attribute control charts assuming that the lifespan of the products under study follows the PFD and WPFD simultaneously when the shape parameter is unknown. We used the average run length (ARL) as the performance criterion to assess the competence of the proposed idea. The results of the ARLs are presented in the form of tables using different parameter values and sample sizes. The paper is concluded as follows.

  • We observed that the WPFD is a more applicable control chart because of its narrower control limits than the PFD control chart.

  • We also observed that the detection ability of the proposed control charts increased with increasing sample size.

  • A real-life application and simulation study were conducted to demonstrate the applicability of the proposed control charts, which showed the outstanding execution of the suggested idea.

  • The proposed control charts can be extended using other sampling schemes, such as repetitive sampling and multiple dependent state sampling, in future work.

  • The proposed chart under WPFD efficiently detected the shift in the process.

  • We suggest further applications of the proposed work in Engineering and Medical Sciences. In addition, more probability distributions could be used to construct the ACC under TTLT.

Comments on this article Comments (0)

Version 1
VERSION 1 PUBLISHED 27 Jun 2024
Comment
Author details Author details
Competing interests
Grant information
Copyright
Download
 
Export To
metrics
Views Downloads
F1000Research - -
PubMed Central
Data from PMC are received and updated monthly.
- -
Citations
CITE
how to cite this article
Zaka A and Jabeen R. Computational Evaluation For The Survival Times Of Patients Under Truncated Life Tests: Simulation And Application [version 1; peer review: 3 approved with reservations]. F1000Research 2024, 13:700 (https://doi.org/10.12688/f1000research.146106.1)
NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.
track
receive updates on this article
Track an article to receive email alerts on any updates to this article.

Open Peer Review

Current Reviewer Status: ?
Key to Reviewer Statuses VIEW
ApprovedThe paper is scientifically sound in its current form and only minor, if any, improvements are suggested
Approved with reservations A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.
Not approvedFundamental flaws in the paper seriously undermine the findings and conclusions
Version 1
VERSION 1
PUBLISHED 27 Jun 2024
Views
2
Cite
Reviewer Report 29 Aug 2024
Shakila Bashir, Forman Christian College, Lahore, Punjab, Pakistan 
Approved with Reservations
VIEWS 2
The authors utilized attribute control charts for two positively skewed distributions—the power function distribution (PFD) and the weighted power function distribution (PWFD). They compared the performance of both types of control charts, presenting their findings in the form of tables. ... Continue reading
CITE
CITE
HOW TO CITE THIS REPORT
Bashir S. Reviewer Report For: Computational Evaluation For The Survival Times Of Patients Under Truncated Life Tests: Simulation And Application [version 1; peer review: 3 approved with reservations]. F1000Research 2024, 13:700 (https://doi.org/10.5256/f1000research.160146.r297249)
NOTE: it is important to ensure the information in square brackets after the title is included in all citations of this article.
Views
0
Cite
Reviewer Report 29 Aug 2024
Alaa El-Alosey, Tanta University, Tanta, Egypt 
Approved with Reservations
VIEWS 0
I have some comments to improve the current version of the paper. The author should addressed the following comments carefully.

1.   In the introduction section, the authors have already done a good job on the literature ... Continue reading
CITE
CITE
HOW TO CITE THIS REPORT
El-Alosey A. Reviewer Report For: Computational Evaluation For The Survival Times Of Patients Under Truncated Life Tests: Simulation And Application [version 1; peer review: 3 approved with reservations]. F1000Research 2024, 13:700 (https://doi.org/10.5256/f1000research.160146.r309911)
NOTE: it is important to ensure the information in square brackets after the title is included in all citations of this article.
Views
4
Cite
Reviewer Report 27 Jul 2024
Ahmed Z. Afify, Benha University, Banha, Al Qalyubia Governorate, Egypt 
Approved with Reservations
VIEWS 4
I have some comments to improve the current version of the paper. The author should addressed the following comments carefully.

1.   In the introduction section, the authors have already done a good job on the literature ... Continue reading
CITE
CITE
HOW TO CITE THIS REPORT
Afify AZ. Reviewer Report For: Computational Evaluation For The Survival Times Of Patients Under Truncated Life Tests: Simulation And Application [version 1; peer review: 3 approved with reservations]. F1000Research 2024, 13:700 (https://doi.org/10.5256/f1000research.160146.r297242)
NOTE: it is important to ensure the information in square brackets after the title is included in all citations of this article.

Comments on this article Comments (0)

Version 1
VERSION 1 PUBLISHED 27 Jun 2024
Comment
Alongside their report, reviewers assign a status to the article:
Approved - the paper is scientifically sound in its current form and only minor, if any, improvements are suggested
Approved with reservations - A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.
Not approved - fundamental flaws in the paper seriously undermine the findings and conclusions
Sign In
If you've forgotten your password, please enter your email address below and we'll send you instructions on how to reset your password.

The email address should be the one you originally registered with F1000.

Email address not valid, please try again

You registered with F1000 via Google, so we cannot reset your password.

To sign in, please click here.

If you still need help with your Google account password, please click here.

You registered with F1000 via Facebook, so we cannot reset your password.

To sign in, please click here.

If you still need help with your Facebook account password, please click here.

Code not correct, please try again
Email us for further assistance.
Server error, please try again.