Biomedical Engineering Reference
In-Depth Information
8
<
(1;C]
if T C
1. Current status data or Case 1 (C1) data, where I =
:
(C;1) if T > C;
where C is a censoring variable. Consider an animal sacrifice study in
which a laboratory animal must be dissected to check whether a tumor
has developed. In this case, T is the onset of tumor and C is the time of
the dissection, and we only can infer at the time of dissection whether
the tumor is present or has not yet developed. A C1 data example is
mentioned in Ayer et al. (1995).
8
<
(1;U]
if T U
2. Case 2 (C2) data, where I =
where (U;V ) is
(V;1)
if T > V
:
(U;V ]
if U < T V;
a censoring vector. In medical follow-up studies, each patient has several
follow-ups and the event of interest is only known to take place either
before the first follow-up U, or between two consecutive follow-ups U
and V , or after the last one, say V . A C2 data example is given in
Becker and Melbye (1991).
3. Mixed IC (MIC) data. MIC data are mixture of exact observations [T;T]
and C2 data. An example is The National Longitudinal Survey of Youth
1979{1998 (NLSY). The 1979{1998 cross-sectional and supplemental
samples consist of 11,774 respondents, who were between the ages of
14 and 22 in 1979. Interviews were conducted yearly from 1979 through
1994; since then data were recorded bi-annually. One entry is the age
at first marriage. There are interval-censored, exact, right-censored, and
left-censored observations in the NLSY data.
For these three types of data, we only know that the survival time T
belongs to the interval I with endpoints L and R. In the literature, if L < R,
people either assume L < T R (see Groeneboom and Wellner (1992)), or
L < T < R (see Li et al. (1997)), or L T R (see Peto (1973)). There
is only a minor conceivable difference in the estimation and there is little
Search WWH ::
Custom Search