no one can help me?
Hello, I have to evaluate the treatment efficacy in a cohort of people over a year (so I have 4 groups of data, one for each 3 months) considering a single variable. This variable is ordinal (goes from 0 to 9) and of course the 4 groups of data are referred to the same persons. Which statistic test shall I use? I thought about t-test (but which one?) but maybe theres something more appropriate.
I use SPSS.
thank you very much
example
followup t0 t1 t2 t3
.
person1: 2 4 4 5
person2: 1 3 2 5
person3: 0 0 1 1
....
p.s. I could also perform this statistic using a three year database, so the groups of data will be much more
no one can help me?
hi,
what is exactly the question you need to answer with the test?
regards
rogojel
Rerun (09-17-2013)
I can compare the cohort of person undergoing the threatment with another cohort undergoing a different one. basically the dataset is huge and some of them chose a treatment, some another, the rest went for placebo. For all of them the initial conditions were the same and there was a three months follow-up.
I need a test that makes me answer the question: which treatment is the best?
thank you so much
Could you describe your dependent variable in more detail? Maybe it could
be treated as an interval scaled measure.
How large are your 3 samples? And is it correct to assume that no randomization
was used to allocate subjects to groups?
With kind regards
K.
Last edited by Karabiner; 09-18-2013 at 03:14 AM. Reason: typo corrected
Rerun (09-17-2013)
the samples contain more than 300 records each
the dependent variable is a score that is related to an effect. it goes from 0 to 9
yes, unfortunately no randomization has been done
thank you so much
I am afraid that this is not very informative. Is it a rating scale? A physical measurement?the dependent variable is a score that is related to an effect. it goes from 0 to 9
Was it already used and validated in other studies? How was it treated there, always as
ordinal scaled?
A repeated measures (mixed) ANOVA would be straightforward,unfortunately no randomization has been done
with time of measurement as within-subjects factor and group
as between subjects factor. But this assumes an interval scaled
dependent variable.
With kind regards
K.
Rerun (09-18-2013)
your assumption was right about the variable. So the anova fits this dataset better? and what type of?
It was described in my last post.
Tweet |