Summary Additional Literature ECP
Week 2. A Framework for Investigating Event Occurrence
- Researchers who want to study event occurrence must learn how to think about their data
in new and unfamiliar ways
- Researchers use survival analysis in a variety of context that shares a common
characteristic: interest centers on describing whether events occur or when events occur.
- If your research question include either word whether or when, you probably need to use
the survival method
- To determine whether a research question lends itself to survival analysis, you must also
examine a study’s methodological features. Even though these studies are set in different
disciplines, they share three methodological features that make them a candidate for
survival analysis
o Target event, whose occurrence is being studied
o Beginning of time, an initial starting point when no one under study has yet
experienced the target event
o Metric for clocking time, a meaningful scale in which event occurrence is recorded
- The only requirement for survival analysis is that, in any particular research setting, the
states be both mutually exclusive (nonoverlapping) and exhaustive (of all possible states)
- In most application each person can occupy only two possible states: drinking or
abstinent. In other applications, each individual can occupy three or more possible states
(being in school, dropping out, or graduating). Some states can be occupied only once in a
lifetime – first word, first step, puberty, etc.
- States must be defined precisely, with clear guidance indicating the behaviors, responses,
or scores constituting each state. Empirical researchers often find this the most difficult
requirement to meet
- The ‘beginning of time’ is a moment when everyone in the population occupies one, and
only one, of the possible states
- To identify the ‘beginning of time’ in a given study, imagine placing everyone in the
population on a time-line, an axis with the ‘beginning of time’ at one end and the last
moment when event occurrence could be observed at the other (birth)
- Once you identify the beginning of time, you must select the units in which you will record
its passage. Sometimes you can record time using thin precise units (e.g. days). Time
should be recorded in the smallest possible units relevant to the process under study
- Despite the obvious advantages of measuring time as precisely as possible, many
researchers find themselves with discrete time data, because
o Some events can occur only at discrete points in time (graduating)
o Although some events can theoretically occur across a wide range of times, many
individuals do not experience them this way. Although employees with term
contracts might be able to quit at any time, most leave when their contract expires
o Data collection constraints often force researchers to use intervals to record the
passage of time
- What should you do with the data from those individuals who do not experience the target
event during the period of data collection? They are labeled censored observations
o The amount of censoring in a study is related to two factors: (1) the rate at which
events occur; and (2) the length of data collection. Censoring can be minimized by
design, but it can rarely be eradicated
- Different types of censoring
Week 2. A Framework for Investigating Event Occurrence
- Researchers who want to study event occurrence must learn how to think about their data
in new and unfamiliar ways
- Researchers use survival analysis in a variety of context that shares a common
characteristic: interest centers on describing whether events occur or when events occur.
- If your research question include either word whether or when, you probably need to use
the survival method
- To determine whether a research question lends itself to survival analysis, you must also
examine a study’s methodological features. Even though these studies are set in different
disciplines, they share three methodological features that make them a candidate for
survival analysis
o Target event, whose occurrence is being studied
o Beginning of time, an initial starting point when no one under study has yet
experienced the target event
o Metric for clocking time, a meaningful scale in which event occurrence is recorded
- The only requirement for survival analysis is that, in any particular research setting, the
states be both mutually exclusive (nonoverlapping) and exhaustive (of all possible states)
- In most application each person can occupy only two possible states: drinking or
abstinent. In other applications, each individual can occupy three or more possible states
(being in school, dropping out, or graduating). Some states can be occupied only once in a
lifetime – first word, first step, puberty, etc.
- States must be defined precisely, with clear guidance indicating the behaviors, responses,
or scores constituting each state. Empirical researchers often find this the most difficult
requirement to meet
- The ‘beginning of time’ is a moment when everyone in the population occupies one, and
only one, of the possible states
- To identify the ‘beginning of time’ in a given study, imagine placing everyone in the
population on a time-line, an axis with the ‘beginning of time’ at one end and the last
moment when event occurrence could be observed at the other (birth)
- Once you identify the beginning of time, you must select the units in which you will record
its passage. Sometimes you can record time using thin precise units (e.g. days). Time
should be recorded in the smallest possible units relevant to the process under study
- Despite the obvious advantages of measuring time as precisely as possible, many
researchers find themselves with discrete time data, because
o Some events can occur only at discrete points in time (graduating)
o Although some events can theoretically occur across a wide range of times, many
individuals do not experience them this way. Although employees with term
contracts might be able to quit at any time, most leave when their contract expires
o Data collection constraints often force researchers to use intervals to record the
passage of time
- What should you do with the data from those individuals who do not experience the target
event during the period of data collection? They are labeled censored observations
o The amount of censoring in a study is related to two factors: (1) the rate at which
events occur; and (2) the length of data collection. Censoring can be minimized by
design, but it can rarely be eradicated
- Different types of censoring
