The name is absent

27

2.5 Literature review of the dependent censoring
and sensitivity analysis

The coexistence of dependent and independent censoring has been investigated by
various researchers. Scharfstein et al (2001), Scharfstein and Robins (2002), and
Tsiatis (1975) have shown that extra information regarding the censoring must be
collected in order to precisely measure the impact due to dependent (informative)
censoring.

There is rich literature on sensitivity analysis. In the setting of survival analysis,
see, for example, Slud (1992) and references therein. Troxel, Ma, and Heitjan (2004)
proposed an index of local sensitivity to non-ignorability. Zhang and Heitjan (2006),
Siannis, Copas, and Lu (2005) and Siannis (2004) used parametric survival models
to do sensitivity analysis.

However, none of these methods can be used to do a sensitivity analysis for the
most widely used Cox (1972) proportional hazards models. Park, Tian and Wei
(2006) proposed a sensitivity analysis method in the nonparametric setting without
covariates. But the method can hardly be extended to general regression problems be-
cause it is incapable of identifying values of the regression parameters with dependent
censoring under a nonparametric setting.

The latest work in this area was by Peng and Fine (2007). They did regression
analysis for semi-competing risk data. It is not clear whether their estimation method

<<   37   38   39   40   41   42   43   >>

More intriguing information