Dependent Data?

[jrb43]

I hope this is the appropriate forum for this:

I would be very grateful if anyone were able to advise on this; my background in statistics is relatively limited. I’m trying to compare the use of different antibiotics on a ward either side of a change in policy which should have discouraged the use of A/B/C and potentially increased the use of D/E/F.

The problem I have is that instead of usage data, which is very difficult to collect, I’ve got supply data for each month of a two year period (one year either side of the policy change). What I’ve noticed is that when you simply plot antibiotic supply against month, the result is almost a sine wave! This makes me think that the data points are not entirely independent due to a scenario something like this:

January – Lots of antibiotic A were used
February – cupboard appeared bare so lots of antibiotic A were ordered
March – Quite a lot of antibiotic A left in cupboard, less ordered
April – etc. etc.

I’m not at all sure what I’m “allowed to do” with the data? I've applied non-parametric testing and demonstrated significance but am concerned that that is not appropriate.

Any help gratefully received; obviously significant help will be credited at publication.

[bugman]

Hey there,

I saw this post and thought that this senario looks familar...

your data, by analogy follows similar patterns that we deal with in water quality. Temperature for example follows strong diurnal and seasonal patterns that are often modelled used sine or cosine dummy variables. The pre- post policy change would also suggest using a intervention time series model - which is in the family of BACI designs.

Might I suggest tracking down the attached reference as a starting point. I have found that it is one of the most referenced articles in the is field and might be a good starting point.

Although not entirely related to your field, I would also suggest searching for articles on trend analysis, intervention analysis - epecially in water quality. This should give you a nudge in right direction.

Phil

[jrb43]

Bugman,

This has been extremely useful; thank you. It's given me plenty to read and get to grips with and I feel much happier analysing my data now!