When data is collected on failure or survival a list of times is
obtained. Some of the times are failure times and others are the times
at which the subject left the experiment. These times both give
information about the performance of the system. The two types will be
referred to as failure and censoring times (cf. Smith section 5). * A
censoring time, t, gives less information than a failure time, for it is
* known only that the item survived past t and not when it failed. The
data is tn and of censoring thus collected as a list of failure times t,
. . ., l * * * times t, t, . . ., t - 1 z m 2. 2. Classical methods
The failure times are assumed to follow a parametric distribution F(t;B)
with and reliability R(t;B). There are several methods of estimating
density f(t;B) the parameter B based only on the data in the sample
without any prior assumptions about B. The availability of powerful
computers and software packages has made the method of maximum
likelihood the most popular. Descriptions of most methods can be found
in the book by Mann, Schafer and Singpurwalla (1974). In general the
method of maximum likelihood is the most useful of the classical
approaches. The likelihood approach is based on constructing the joint
probability distrilmtion or density for a sample.