Survivor Function Analysis & Distribution Fitting for Failure Times & Survival Data, Exercises of Mathematical Statistics

Instructions for analyzing survivor functions and fitting various distributions to failure times and survival data from experiments and patient studies. Topics include maximum likelihood estimation, confidence intervals, likelihood ratio tests, bic and aic procedures, and kaplan-meier estimates. The data sets involve high-speed turbine engines, melanoma patients, polyethylene cable insulation, electrical appliances, and breast cancer patients.

Typology: Exercises

2012/2013

Uploaded on 01/11/2013

bulla.baba
bulla.baba 🇮🇳

5

(7)

87 documents

1 / 2

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
1. An experiment was designed to compare the performances of high-speed
turbine engine made out of five different compounds. The exp eriment tested 10
bearing of each type. The time to fatigue failure are given in units of million of
cycles in my web site.
(a) Assuming that the failure time in each sample came from a Weibull distri-
bution, obtain the maximum likelihood estimates of parameters and find 95%
confidence interval for the tenth percentile of each distribution.
(b) Give the plot of five survivor functions on the same axes. carry out a
comparison of the five survivor functions. Test the equality of five survivor
functions.
(c) Investigate whether a log-normal distribution also fits the data by considering
a generalized gamma model. Compare confidence interval for tenth percentile
under a log-normal model with those in part (a)
2. The survival times of thirty-one patients with advanced melanoma treated
with combined chemotherapy, immunotherapy, and hormonal terapy are given
in my web page, where + indicate observation is censored.
(a) Make a probability plot for each of the following distributions: exponential,
Weibull, lognormal, and log-logistic distribution. Which distribution provides a
reasonable fit to the data? Estimate the parameter of the distribution chosen.
(b) Make the Cox-Snell residual plot for each of the following distributions: ex-
ponential, Weibull, lognormal, log-logistic, and generalized gamma distribution.
Which distribution provides a reasonable fit to the data?
(c) Obtain the log-likelihoods for exponential, Weibull, lognormal, and general-
ized gamma distribution. Perform the likelihood ratio test and select the best
distribution among these four distributions.
(d) Use BIC and AIC procedure to select the best distribution among the four
distributions in part (c) plus the log-logistic distribution.
(e) Compare the results obtained in parts (a)(d). In addition, compare
the maximum likelihood estimates of the parameters with those obtained by
graphical methods.
1
Docsity.com
pf2

Partial preview of the text

Download Survivor Function Analysis & Distribution Fitting for Failure Times & Survival Data and more Exercises Mathematical Statistics in PDF only on Docsity!

  1. An experiment was designed to compare the performances of high-speed turbine engine made out of five different compounds. The experiment tested 10 bearing of each type. The time to fatigue failure are given in units of million of cycles in my web site.

(a) Assuming that the failure time in each sample came from a Weibull distri- bution, obtain the maximum likelihood estimates of parameters and find 95% confidence interval for the tenth percentile of each distribution.

(b) Give the plot of five survivor functions on the same axes. carry out a comparison of the five survivor functions. Test the equality of five survivor functions.

(c) Investigate whether a log-normal distribution also fits the data by considering a generalized gamma model. Compare confidence interval for tenth percentile under a log-normal model with those in part (a)

  1. The survival times of thirty-one patients with advanced melanoma treated with combined chemotherapy, immunotherapy, and hormonal terapy are given in my web page, where + indicate observation is censored.

(a) Make a probability plot for each of the following distributions: exponential, Weibull, lognormal, and log-logistic distribution. Which distribution provides a reasonable fit to the data? Estimate the parameter of the distribution chosen.

(b) Make the Cox-Snell residual plot for each of the following distributions: ex- ponential, Weibull, lognormal, log-logistic, and generalized gamma distribution. Which distribution provides a reasonable fit to the data?

(c) Obtain the log-likelihoods for exponential, Weibull, lognormal, and general- ized gamma distribution. Perform the likelihood ratio test and select the best distribution among these four distributions.

(d) Use BIC and AIC procedure to select the best distribution among the four distributions in part (c) plus the log-logistic distribution.

(e) Compare the results obtained in parts (a) − (d). In addition, compare the maximum likelihood estimates of the parameters with those obtained by graphical methods.

Docsity.com

  1. Failure times for two types of polyethylene cable insulation, obtained from an accelerated life test are recorded. Of the 10 specimens of each type tested, 9 failed. Ordered failure time, in hours, are given in my web site. The last time in each case is a censoring time. Assume that the failure times for each type have a Weibull distribution.

(a) Compare two survivor functions by likelihood ratio test.

(b) Assess the possible superiority of Type II insulation.

  1. The number of cycles to failure for a group of 60 electrical appliance in a life test is given in my web site. The failure time have been ordered for convenience.

(a) Give the Kaplan-Meier estimate and plot of survivor function.

(b) The data suggest that the hazard function might have two components, one consisting of a fairly small portion of the distribution and giving small failure times, and one giving a wide range of larger failure time. Therefore, we might consider the following mixture model

f (t) = pf 1 (t) + (1 − p)f 2 (t)

and consequently S(t) = pS 1 (t) + (1 − p)S 2 (t)

Consider a mixture of two Weibull componets, with S 1 (t) and S 2 (t) are given by Sj (t) = exp [−(λj t)γj^ ] j = 1, 2 (i) Find the maximum likelihood equations.

(ii) The maximum likelihood estimates of parameters are

ˆp = 0. 137 , ˆλ 1 = 0. 01048 , ˆγ 1 = 1. 66 , ˆλ 2 = 0. 00036 , γˆ 1 = 1. 40.

Find the Cox-Snell residual plot for the mixture model. Does the Weibull mixture agree with observed data?

  1. The time, T , to cosmetic deterioration of the breast was recorded for some patients with breast cancer. The data, given in my web page, are right, left or interval censored data. Is there a difference in the distribution of T for women who received radiation therapy (R) only versus a combination of radiation and chemotherapy (RC)? To answer the question, obtain and compare (parametric and non-parametric) estimates of the survivor function S(t) for each group.
  2. Show that the Cox-Snell residuals defined by Ri = − ln S(Ti) follow the unit exponential distribution with density f (r) = exp(−r)

Docsity.com