STATS MAKE ME CRY

Hi,

I'm currently trying to work out the best method of analysis for my data, and would really appreciate some advice.

The data that I am analysing was collected longitudinally over a four year period with patient - companion pairs. Patients rated themselves, and companions rated the patients. At baseline, patients were at different levels of illness severity, and over the four year course, I have some measures of change / re-staging of disease. (Sorry, I should state, the patients have a degenerative disease, so illness severity worsens over time at variable rates).

I'm looking at differences in their ratings to explore patients' awareness of problems. I'm keen to examine two main things in the analysis:

a) does discrepancy between participant - companion rating relate to severity of disease?

b) do patterns of individual participant - companion pairs vary over the four year period (for example, are there particular pairs who are always hugely discrepant in their ratings / always rate similarly)

So, I have longitudinal data (4 years) and a single outcome variable (participant-companion rater difference). The participants start at different levels of disease severity (random intercepts?), and I have a few different measures of this that I could potentially use. When plotted they also seem to show different patterns of rating discrepancy over the four year period (random slopes?) that as far as I can tell look kind of 'cubic' in nature.

Given that I expect the pairs are going to be individually variable and that not all of the individuals will have taken part at every time point I am wondering whether multilevel growth modelling is the best method for this analysis. The problem is that I have very little knowledge of how to really do this in SPSS, even after reading a number of books on the subject and trying it out. My problem seems to be that I don't have an 'intervention' that happens after the 'baseline measure', to predict an 'outcome'. I essentially have now built a dataset that has the following variables:

Participant (a case for each year that the participant took part in the study)
Participant – companion difference score (1-4, some missing data)
Time measured as visit no (1-4)
Time measured in years from baseline
Disease severity score as measured by a disease-specific scale (subject to change each year)
Disease ‘stage’ (subject to change each year)
Site (the data collection took place at several different sites – I’m not sure if this would be a relevant covariate)
Gender
Age

If anyone could give me any ideas about where to start, I would really appreciate it. Just when I think I understand what I’m doing, the examples that I’m reading about don’t really ‘fit’ with my own data, so I lose the thread of how to proceed, and which variables constitute what.

Thanks!