Dear allstatters,
I would greatly appreciate your advice as to whether 2-way repeated
measures analysis is suitable for the following situation:
Interested in the effects of 3 interventions on plantar pressure (for
example), data is gathered on a set of patients who each receive all 3
interventions (randomised in order). For each patient, in each
intervention, data is recorded from multiple footfalls (say 5). Rather
than take the average of the 5 measurements for each patient in each
intervention, would it be valid to use 2-way repeated measures analysis
with two within subjects factors: INTERVENTION (3 levels) and REPEAT (5
levels)?
My main concern is that the examples given in the literature for this
analysis assume that the repeated measures made within each intervention
could be expected to differ. For example, students being asked to detect
3 different shapes (circle, triangle, square) when drawn in outline or
presented as a solid. The different shapes here represent the equivalent
of our 3 interventions, and the solidity variable is the equivalent of
our 5 footfalls. However, although solidity is a repeated measure within
subjects, the two levels represent measurements made on different items
and one could reasonably expect there to be a degree of variation
between the two. We have no reason to expect such a degree of variation
between footfalls, and would rather expect the results to be internally
correlated. Is 2-way repeated measures the right technique to use, and
can we correct for this internal correlation, or should we be using
different analysis altogether? We are not so much interested in the
effects of the different footfalls (as we do not expect there to be
any), rather we are concerned that by taking an average we may be losing
considerable amounts of information.
Thanks for you time
Liz Hensor
Dr Elizabeth M A Hensor PhD BSc (Hons)
Data Analyst
Academic Unit of Musculoskeletal and Rehabilitation Medicine
36 Clarendon Road
Leeds
West Yorkshire
LS2 9NZ
Tel: +44 (0) 113 3434933
Fax: +44 (0) 113 2430366
[log in to unmask]
|