Transitions which appear to occur immediately after
interviews
Two main sources
Either a true new spell is reported with an erroneously
early date (i.e., starts before last year's interview)
Or state information changes (e.g., occupational code changes
due to trivial or erroneous re-description of same job)
In the former case, the measurement error is in the date, but the
transition is real (one strategy is to randomise the date)
In the latter case, the transition may be real (error in ``same
state'' claim) or spurious (error in state information): a need
to judge cases on their merits.
It is necessary to provide as much information as possible
about the seam effect to the end user
This is the reason for newpan's somewhat complex structure
Somewhat more effort required of user
Much more structural information retained
Key concept: division into ``splits'' and ``episodes''
Splits are periods where the information is coming from a
given source, e.g., CINDRESP or the third record in
FJOBHIST and are represented as single records
Episodes are substantive employment status spells and
represented as one or more consecutive splits.
Splits are directly related to the main BHPS database
(given the particular algorithms used)
Which sets of splits are taken to constitute episodes depends
(in part) on a user-level judgement about seams
