Well, if your "diagnostic criteria" variable can't be treated as an
interval scale, that completely changes how you must model the data. I
don't know about possible approaches for specifically utilizing the
ordinal (ranked) nature of that variable, but one approach would be to
treat the variable as categorical, and model its 4 levels in the same
way that you would model something like group. Then use an F test
across the 4 levels to see if there is an overall relationship with your
dependent variable.
cheers,
-MH
On Thu, 2011-05-19 at 11:33 -0400, Rajagopalan, Venkateswaran wrote:
> Dear Michael,
>
> Thanks so much for clarifying now I understand better. One last question though as you mentioned my "diagnostic criteria" are actually coding a categorical scale (in my case I can say that it is an ordinal variable but not definitely an interval data) so in this case how can I study the relationship between GM volume changes and this diagnostic criteria.
>
> Thanks a lot
>
> Venkat
>
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library on behalf of Michael Harms
> Sent: Thu 5/19/2011 11:13 AM
> To: [log in to unmask]
> Subject: Re: [FSL] Contrast matrix-FSL VBM
>
> Ok, I guess I read your email incorrectly then.
>
> Each group EV should contains 1's indicating the members of that group,
> and 0's otherwise. Your use of "diagnostic criteria" as EV5 is ok
> PROVIDED that the values of that EV constitute an interval scale. If
> the values of EV5 are actually coding a categorial scale (see
> http://en.wikipedia.org/wiki/Level_of_measurement), then the coding of
> EV5 isn't meaningful. If EV5 is indeed a meaningful interval scale,
> then the contrast
> 0 0 0 0 1
> would tell you whether there is a significant (+ve) relationship across
> all subjects between "diagnostic criteria" and your dependent variable
> while controlling for group status.
>
> The contrasts
> Ev1 Ev2 Ev3 Ev4 Ev5
> 1 0 0 0 1
> 0 1 0 0 1
> 0 0 1 0 1
> 0 0 0 1 1
>
> are not testing anything meaningful.
>
> cheers,
> -MH
>
>
> On Thu, 2011-05-19 at 10:51 -0400, Rajagopalan, Venkateswaran wrote:
> > Dear Michael,
> >
> > Thanks I am able understand your answer to some extent. Kindly help me with the following
> >
> > 4 group EVs, and then a 5th EV in which you code an integer "diagnostic criteria" in which each integer corresponds to a different group
> >
> > Sorry I am not able to understand here "in which each integer corresponds to a different group"
> >
> > Here is what I have for each subject in each group I have diagnostic criteria score as given below
> >
> > patient group 1(EV1) patient group 2 (EV2) ..................diagnostic criteria (Ev5)
> > subject 1 GM vol 1
> > Subject 2 GM vol 2
> > subject 3 GM vol 4
> > . subject 1 GM vol 3
> > subject 2 GM vol 1
> > . . .
> > . . .
> > . . .
> > .
> > So If I want to correlate GM volume changes with my diagnostic criteria, how can I design my EVs and contrast matrix. I didn't classify my patients groups based on diagnostic criteria but I classified them based on their clinical phenotype evaluated using clinical signs and symptoms but I am interested to see is there is correlation between GM volume changes and their diagnostic criteria.
> >
> > Also can you kindly tell me whether my design and contrast matrix is correct for my other clinical criteria disease duration
> >
> > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 0 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > >
> > > or
> > > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 1 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > > 0 1 0 0 1
> > > 0 0 1 0 1
> > > 0 0 0 1 1
> >
> > Thanks
> >
> > Venkat
> > ________________________________
> >
> > From: FSL - FMRIB's Software Library on behalf of Michael Harms
> > Sent: Thu 5/19/2011 10:22 AM
> > To: [log in to unmask]
> > Subject: Re: [FSL] Contrast matrix-FSL VBM
> >
> >
> >
> > Hi Venkat,
> >
> > If I'm reading your email correctly, it sounds as if you have a design
> > with 4 group EVs, and then a 5th EV in which you code an integer
> > "diagnostic criteria" in which each integer corresponds to a different
> > group. If so, then that design doesn't make sense since "diagnostic
> > criteria" is already being modeled by your 4 group EVs (and you can't
> > model a "class" or "categorical" variable involving more than two levels
> > with a single EV anyway).
> >
> > cheers,
> > -MH
> >
> > On Thu, 2011-05-19 at 09:58 -0400, Rajagopalan, Venkateswaran wrote:
> > > Dear All,
> > >
> > > I read FSL archives and Steve also clarified me with some of my
> > > questions but I am not sure about the following contrast matrix which
> > > I designed to correlate between clinical measures and GM volume
> > > changes. My question here is does the GM volume changes correlate with
> > > clinical measures
> > >
> > > I have 4 groups of patient groups (let us say group1, group2, group3
> > > and group4) and I have clinical measures disease duration (in number
> > > of months), diagnostic criteria which varies from 1-4 and disease
> > > progression
> > >
> > > As per Steve's suggestion I tried to design a model for each clinical
> > > measure for instance for diagnostic criteria of 1-4. My question here
> > > does the GM volume change correlate diagnostic criteria. I have 5 EVs
> > > with Ev1..Ev4 for patient groups (i.e. their GM volume changes) and
> > > Ev5 is the dignostic criteria.Now I would like to know which of the
> > > below contrast matrix is correct
> > >
> > > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 0 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > >
> > > or
> > > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 1 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > > 0 1 0 0 1
> > > 0 0 1 0 1
> > > 0 0 0 1 1
> > >
> > > OR do I need to model by grouping my patients (my current method of
> > > grouping patients is based on their different clinical phenotypes)
> > > based on diagnostic criteria i.e. group1 (diagnostic criteria =1),
> > > group2 (diagnostic criteria =2)...
> > >
> > >
> > > Similarly for my clinical measure of disease duration
> > > 5 EVs with Ev1..Ev4 for patient groups (i.e. their GM volume changes)
> > > and Ev5 is the disease duration
> > >
> > > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 0 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > >
> > > or
> > > Ev1 Ev2 Ev3 Ev4 Ev5
> > > 1 0 0 0 1 (for +ve correlation and -1 for -ve correlation)
> > > 0 1 0 0 1
> > > 0 0 1 0 1
> > > 0 0 0 1 1
> > >
> > > Sorry designs similar to these are discussed in FSL archives
> > > unfortunately I am not sure about my contrast matrix.
> > >
> > > Thanks
> > >
> > > Venkat
> > >
> > > ===================================
> > >
> > >
> > > Please consider the environment before printing this e-mail
> > >
> > >
> > >
> > > Cleveland Clinic is ranked one of the top hospitals
> > > in America by U.S.News & World Report (2010).
> > > Visit us online at http://www.clevelandclinic.org <http://www.clevelandclinic.org/> for
> > > a complete listing of our services, staff and
> > > locations.
> > >
> > >
> > > Confidentiality Note: This message is intended for use
> > > only by the individual or entity to which it is addressed
> > > and may contain information that is privileged,
> > > confidential, and exempt from disclosure under applicable
> > > law. If the reader of this message is not the intended
> > > recipient or the employee or agent responsible for
> > > delivering the message to the intended recipient, you are
> > > hereby notified that any dissemination, distribution or
> > > copying of this communication is strictly prohibited. If
> > > you have received this communication in error, please
> > > contact the sender immediately and destroy the material in
> > > its entirety, whether electronic or hard copy. Thank you.
> > >
> >
> >
> >
> >
> > ===================================
> >
> > Please consider the environment before printing this e-mail
> >
> > Cleveland Clinic is ranked one of the top hospitals
> > in America by U.S.News & World Report (2010).
> > Visit us online at http://www.clevelandclinic.org for
> > a complete listing of our services, staff and
> > locations.
> >
> >
> > Confidentiality Note: This message is intended for use
> > only by the individual or entity to which it is addressed
> > and may contain information that is privileged,
> > confidential, and exempt from disclosure under applicable
> > law. If the reader of this message is not the intended
> > recipient or the employee or agent responsible for
> > delivering the message to the intended recipient, you are
> > hereby notified that any dissemination, distribution or
> > copying of this communication is strictly prohibited. If
> > you have received this communication in error, please
> > contact the sender immediately and destroy the material in
> > its entirety, whether electronic or hard copy. Thank you.
>
>
>
> ===================================
>
> Please consider the environment before printing this e-mail
>
> Cleveland Clinic is ranked one of the top hospitals
> in America by U.S.News & World Report (2010).
> Visit us online at http://www.clevelandclinic.org for
> a complete listing of our services, staff and
> locations.
>
>
> Confidentiality Note: This message is intended for use
> only by the individual or entity to which it is addressed
> and may contain information that is privileged,
> confidential, and exempt from disclosure under applicable
> law. If the reader of this message is not the intended
> recipient or the employee or agent responsible for
> delivering the message to the intended recipient, you are
> hereby notified that any dissemination, distribution or
> copying of this communication is strictly prohibited. If
> you have received this communication in error, please
> contact the sender immediately and destroy the material in
> its entirety, whether electronic or hard copy. Thank you.
|