Parametric survival models with interval censored data in determining prognostic factors of patients of lung cancer

• Main Author: Muhamad Jamil, Siti Afiqah
• Format: Thesis
• Language: English; English; English
• Published: 2020
• Subjects: TK7800-8360 Electronics
• Online Access: http://eprints.uthm.edu.my/1172/

In clinical trials, biological research and medical studies that involved periodically follow-ups, it is predominently to have censored data. The censored data can be either left, right or interval censored where it reflects on the uncertainty of survival time until an event occur. Survival analysis can accommodates both fixed and time varying covariates with the presence of censored data. The survival time of parametric distribution of Weibull, exponential and log-logistic were derived by using the inverse cumulative distribution function with the hazard and survival function. Standard estimation values such as, the mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and mean asolute percentage error (MAPE) were employed in comparing different distributions and number of sample sizes. Besides, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Corrected Akaike Information Criterion (AICC) been evaluated in finding the best fit model towards the survival time of lung cancer. Thus, the exponential model was found to be the most reliable with censored and fixed covariate of simulation and lung cancer data while the log-logistic appeared to be practically more stable than Weibull in estimating the censored with varying effect covariate. Meanwhile, the prognostic factors that were significant involved the types of lung cancer, gender and some other interactions. Somehow, increased number of sample either in simulation or bootstrap makes the results to be approximately more reliable as the biases decreased.