Hello,
I was wondering if anyone could offer me advice. I would like to know if it is
possible to treat repeated measures data as independent in some cases.
I want to plot length of time at university with Measure X. If the outcome
of this scatter plot was linear, I would then like to look at the
correlation between Measure X and length of time at university (If it's
curvelinear then I'd think of some other type of regression analysis I suppose).
My time points are 0 months, 6 months, 12 months, 18 months, 24 months
and 30 months.
For the sake of illustration let's say that my other measure (X) is
Attitude towards cannabis.
I have asked students at various stages in their university career to rate
their attitude towards cannabis on a Likert scale. I have 867 'measures' Now
if this was a simple cross section, I would have no trouble, because my 867
'measures' would be 867different people. But therein lies my problem. I don't
actually have 867 different people. I have 494 different people, some of who
answered the same question 4 different time points (at different stages in
their university 'career'). I will clarify this further. I have four separate groups of
participants, groups (a), (b), (c), and (d).
at 0 months, 136 people from group (a) responded, along with 114 people
from group (d) TOTAL = 250
at 6 months, 56 people from group (a) responded, along with 40 people from
group (d) TOTAL = 96
at 12 months, 120 people from group (b) responded, along with 51 people
from group (a) TOTAL = 171
at 18 months, 45 people from group (b) responded, along with 45 people from
group (a) TOTAL = 90
at 24 months, 124 people from group (c) responded, along with 47 people
from group (b) TOTAL = 171
at 30 months, 48 people from group (c) responded, along with 41 people from
group (b) TOTAL = 89
Since it's not an 'experimental' design (i.e. I am not manipulating any
variables) there is no reason to expect that those who answered the
questionnaire previously should be any less or more honest about their
attitudes towards cannabis. Could I therefore justify treating the data from
one subject as four 'independent' points in my scatter plot?
Thank you in advance for your help,
Brian
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Brian McMillan
Mother and Infant Research Unit
22 Hyde Terrace,
Leeds LS2 9LN.
Tel: 0113 233 1575
Fax: 0113 244 9730
http://www.leeds.ac.uk/miru
|