Matstat seminarium fredagen 19 oktober kl 13:15 i MH:227


Niels Keiding
Department of Biostatistics, Institute of Public Health,
University of Copenhagen

Sampling frames in event history analysis

In event history analysis individuals are assumed to move between
states, and multi-state models (Andersen et al., 1993, Andersen and
Keiding, 2001) are used for their description. In this presentation I
survey my interests in event histories developing in calendar time,
studied at a cross-section at a particular time. The Lexis diagram
(Keiding, 1990, 2000) is helpful here.

A first example is confirmatory analysis of a possible chance finding at
an interim analysis of a clinical trial with staggered entry, obtaining
by reusing, with delayed entry, the survivors from the interim analysis
(Keiding et al., 1987, Parner and Keiding, 2001).

Age-specific generalizations of the epidemiological relationships
between incidence, prevalence and duration may be formalized in this
framework (Keiding, 1991, Lund, 2000). The individuals in the prevalent
sample may be followed up using standard delayed entry methodology
(Keiding and Gill, 1990, Keiding, 1992, Lund, 2000). Furthermore, it is
sometimes possible to estimate incidence retrospectively from the
information about the survivors in the prevalent sample, provided
reliable supplementary information on survival is available to obtain
Horvitz-Thompson type weights (Keiding et al., 1989, Ogata et al.,
2000). Interaction between life history events, allowing non-symmetric
dependence concepts, may also be assessed from the retrospective
information in a cross-sectional sample (Aalen et al., 1980).

If only the current status data are available, age-specific incidence
may be estimated assuming time-homogeneity (Keiding, 1991, Keiding et
al., 1996). A recent variant is current duration data: derive from the
waiting time experienced so far an estimate of the distribution of time
to pregnancy (Keiding et al., 1999).