A thought occurs. Longitudinal data allows change to be studied within individuals was well as within populations. This is ‘a good thing’. You can look at whether a training programme affected an individual’s employment status, for example. Or whether breakfast altered their feelings of hunger. This does require that you have measured the same phenomenon at a number of time points.
What I have is a dataset (NCDS) which asks related-but-not-identical questions at each time point (aka wave). It also uses extremely broad scales. That is, it asks people to tick ‘often/sometimes/never’ rather than asking them to tick ‘1-2 hours/3-4 hours/5-6 hours/7-8 hours/9-10 hours/more than 10 hours per week’. It’s therefore a bit tricky to draw a nice trajectory for an individual respondent. I’m now back to wondering whether I should be trying to explain responses in 2008 using an individual’s previous responses, rather than trying to explain a lifetime trajectory…