|
Abstract:
|
Barnett and Cohen (2000, JCGS) extended the definition of the histogram to accommodate right-censored data. We elaborate their histogram. We first extend it to left-truncated and right-censored data, which often occur in medical studies. Next, we introduce a new parametric censored histogram based on log-location and scale distributions, which are commonly used for lifetime data modeling. Two measures of discrepancy between Kaplan-Meier estimator-based histogram and the proposed parametric censored histogram are automatically computed for each of the distributions to identify the true distribution in our simulation study. Results show the parametric censored histogram performs well. Once the true distribution is identified, this censored parametric histogram is also used to detect outliers. Real data examples are shown to illustrate the methods.
|
