| |
 |
|
|
Science Forum Index » Space - Consult Forum » What kind of stats should I use?...
Page 1 of 1
|
| Author |
Message |
| ... |
 Posted: Tue Jul 22, 2008 8:48 am |
|
|
|
Guest
|
Hi,
Here is the design of the experiment I want to analyze :
We have 6-10 patients per therapy group of a couple of weeks, each
group being in a different location and therapists changing by
location. For each group, there is 2 therapists. We have a measure
of symptoms, symptoms being measured at baseline, during treatment and
two other times after treatment. We also have a measure of working
alliance between patient and each therapist, rated by both the patient
and the therapists 3 times during the therapy. So each patient rates
the alliance with both therapists 3 times during therapy (beginning,
middle and end), and each therapists rates alliance with each patients
3 times during therapy.
What I want to know is : are symptoms improvement (repeated measure)
varying in function of working alliance (which is, again, a repeated
measure)?
What kind of statistical test should I use to answer that question?
Thank you,
Marie-Christine |
|
|
| Back to top |
|
| Ray Koopman... |
| Posted: Tue Jul 29, 2008 9:17 pm |
|
|
|
Guest
|
On Jul 22, 11:48 am, marielaferri... at (no spam) gmail.com wrote:
Quote: Hi,
Here is the design of the experiment I want to analyze :
We have 6-10 patients per therapy group of a couple of weeks, each
group being in a different location and therapists changing by
location. For each group, there is 2 therapists. We have a measure
of symptoms, symptoms being measured at baseline, during treatment and
two other times after treatment. We also have a measure of working
alliance between patient and each therapist, rated by both the patient
and the therapists 3 times during the therapy. So each patient rates
the alliance with both therapists 3 times during therapy (beginning,
middle and end), and each therapists rates alliance with each patients
3 times during therapy.
What I want to know is : are symptoms improvement (repeated measure)
varying in function of working alliance (which is, again, a repeated
measure)?
What kind of statistical test should I use to answer that question?
Thank you,
Marie-Christine
The question appears to call for some kind of within-subject
correlation or regression over time, but I see at least two
problems. First, the small number of time points means that
you won't be able to look at results for individuals, but
only at aggregated results. Second, you should probably make
some correction or adjustment for general drift over time
that is not subject-specific; that is, in anova terms, you
need to correct for time main effects, and possibly therapist
by time interactions. |
|
|
| Back to top |
|
| RichUlrich... |
| Posted: Wed Jul 30, 2008 3:02 pm |
|
|
|
Guest
|
On Wed, 30 Jul 2008 00:17:49 -0700 (PDT), Ray Koopman <koopman at (no spam) sfu.ca>
wrote:
Quote: On Jul 22, 11:48 am, marielaferri... at (no spam) gmail.com wrote:
Hi,
Here is the design of the experiment I want to analyze :
We have 6-10 patients per therapy group of a couple of weeks, each
group being in a different location and therapists changing by
location. For each group, there is 2 therapists. We have a measure
of symptoms, symptoms being measured at baseline, during treatment and
two other times after treatment. We also have a measure of working
alliance between patient and each therapist, rated by both the patient
and the therapists 3 times during the therapy. So each patient rates
the alliance with both therapists 3 times during therapy (beginning,
middle and end), and each therapists rates alliance with each patients
3 times during therapy.
What I want to know is : are symptoms improvement (repeated measure)
varying in function of working alliance (which is, again, a repeated
measure)?
What kind of statistical test should I use to answer that question?
Thank you,
Marie-Christine
The question appears to call for some kind of within-subject
correlation or regression over time, but I see at least two
problems. First, the small number of time points means that
you won't be able to look at results for individuals, but
only at aggregated results. Second, you should probably make
some correction or adjustment for general drift over time
that is not subject-specific; that is, in anova terms, you
need to correct for time main effects, and possibly therapist
by time interactions.
Third -- The hypthesis says that Improv= funct(Alliance),
which probably should be contrasted to Alliance= funct(Improv).
It seems that "Improv" might be taken as a difference, using
the Baseline, before therapy.
At first, I was thinking that the automatic improvements over
time would make it nearly impossible to disentangle the
effects, so that there would have to be a "+" correlation by
artifact alone. Now, however, it occurs to me that it is likely
that the biggest improvement will happen between Baseline
and 1st treatment (at least, if these were acutely-ill subjects
being sent to treatment, and there is a few days lag),
and 1st is where Alliance is apt to be lowest. So, if one is
not careful about what is being measured, and when, there
could be an artifact giving a negative correlation.
I think that I would want to look at plots of individuals scores
on both measures, and design my exploratory analysis to
take into account whatever artifacts can be imagined.
Does the scale bottom-out on Symptoms?
--
Rich Ulrich |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Fri Jan 09, 2009 11:40 pm
|
|