Initial sample is of households but subsequently it is of
individuals in households
we can weight for individuals or households (households have
only cross-sectional weights)
We link within households in each wave by wHID
Within each wave, wHID and wPNO map directly to
PID
Across waves, the pHID, pPNO : qHID,
qPNO is stored in XWAVEID
In recent releases of the BHPS, PID is found everywhere
it might be needed, so use of XWAVEID to work out the
wHID/wPNO corrspondence is less necessary, and
PID variants are more common (e.g., wSPPID to identify
spouse, similarly for certain other relationships)
Enumerated individuals are a superset of respondents
All individuals in the household are enumerated early in
the interview (recorded in ``household grid'', information stored in
wINDALL and wEGOALT)
Non-respondent enumerated individuals consist of children
(under 16) and within-household refusals/non-contacts
Among respondents we distinguish full, proxy and phone
respondents
Proxy and phone responses provide little useful information
at the individual level
But are more important for certain household level
summaries such as income
Important for monitoring non-response, and minimising
resultant bias
Distinction is automatic in practice, proxy/phone
respondents have missing values on most variables so will drop
out of analysis
A key distinction: Permanent versus Temporary Sample Members
PSMs are effectively members of the initial sample, relate
to the reference population (GB in 1991)
TSMs ``accrete'' to the sample and are members by virtue of
shared households with PSMs, and are not really part of the
``panel''
Strictly speaking, if we want to analyse a ``panel'' of
longitudinal respondents whose reference population is well
understood, we exclude the TSMs
When properly weighted the PSMs can be said to approximate
what happens to the 1991 reference population over time
Use of the longitudinal weights wLRWGHT and
wLEWGHT automatically excludes TSMs
The other critical distinction: which sample?
Original ``Essex'' sample
Wave 7 ``ECHP'' boost (intended to oversample low income
households)
Wave 9 Scottish and Welsh Panels: fresh start samples
Wave 11 Northern Irish Panel: fresh start sample
Distinction coded in wMEMORIG
Sets of weights to analyse jointly (cross-sectional weights
only) and separately (both longitudinal and cross-sectional for
new panels)
