4. Time-to-event Models: Example 1#
Consider an exponential survival model:
Let \(\delta_i\) be an indicator to represent censored points. There are \(k\) uncensored points, and \(n-k\) censored points
What is the (frequentist) estimator of \(\lambda\)? If all data points were observed, the MLE of \(\lambda\) would be
Should we ignore censored points? The estimator would be
which is wrong because it is biased to the non-censored data. So, should we consider censored points as observed?
This is also incorrect, because the censored values are likely larger than what was observed.
Now let’s consider the likelihood as
We arrive at this likelihood by taking into account both censored and uncensored points. The non-censored points follow an exponential PDF, while the censored points have probability of the exponential’s complementary CDF.
The MLE of \(\lambda\) is
which is correct!