15-20 possible covariates seem rather extreme. Can you pare down the
list to something smaller based on either a priori principles? Or if you
just want to account for the majority of the variance represented by those
15-20 variables, you could run them through a PCA, and then use the scores
on the first handful of PCs as your covariates.
I don’t think your approach of running a bunch of univariate tests as a
sort of ad hoc model selection procedure is a good option because (1) it
is hard to define a criterion with imaging data (e.g., the different
variables may show significant clusters in differing portions of the
brain), and (2) you aren’t dealing with the collinearity of the covariates
in a systematic manner.
cheers,
-MH
--
Michael Harms, Ph.D.
-----------------------------------------------------------
Conte Center for the Neuroscience of Mental Disorders
Washington University School of Medicine
Department of Psychiatry, Box 8134
660 South Euclid Ave.Tel: 314-747-6173
St. Louis, MO 63110Email: [log in to unmask]
On 5/23/16, 8:21 AM, "FSL - FMRIB's Software Library on behalf of Aa, N.E.
van der" <[log in to unmask] on behalf of [log in to unmask]> wrote:
Dear Michael,
Thank you for your response. I understand that 3 or 4 variables shouldn’t
make that much of a difference, but what if I have 15-20 variables of
which some also are associated with each other? How do I determine which
variable(s) explain(s) most of the variety? Would a “univariate’ model
followed by a multivariate approach as described below be a good option?
> Op 23 mei 2016, om 13:56 heeft Harms, Michael <[log in to unmask]> het
>volgende geschreven:
>
>
> Hi,
> With 150 subjects, I wouldn’t worry too much about including a couple
> extra covariates. The impact on your power will be minimal, and the
> resulting analysis will appropriately control for age, gender, and
>disease.
>
> cheers,
> -MH
>
> --
> Michael Harms, Ph.D.
>
> -----------------------------------------------------------
> Conte Center for the Neuroscience of Mental Disorders
> Washington University School of Medicine
> Department of Psychiatry, Box 8134
> 660 South Euclid Ave.Tel: 314-747-6173
> St. Louis, MO 63110Email: [log in to unmask]
>
>
>
>
> On 5/23/16, 6:43 AM, "FSL - FMRIB's Software Library on behalf of Niek
>van
> der Aa" <[log in to unmask] on behalf of [log in to unmask]> wrote:
>
> Dear FSL users,
>
>
> I’m currently working on a TBSS analysis using a sample size of
> approximately 150 children. I have a number of clinical variables that
>may
> be used as an explanatory variable. Normally, when I would perform a
> linear regression analysis, I would start with univariate testing to
> identify any possible explanatory variables. Any significant variables
> would then be included in multivariate analyses.
>
> Does it make sense to apply the same for TBSS? As age is a wel known
> variable to affect FA in (young) children), so I would like to combine
> each variable with age at scan. So for example, a first ‘univariate
> analysis’ could be gender combined with age at scan, next presence of
> disease (yes/no) and age at scan, etc. All variables that are significant
> would then be included in a multivariate TBSS model.
>
> If the above does make sense, what should I then do with variables that
> show a significant association with FA in the ‘univariate’ analysis but
> show no correlation in multivariate analyses. For example, if gender
> affects FA when analyzed with age only (as mentioned above), but shows no
> effect on FA when combined with presence of disease (yes/no), number of
> medicine and age at scan. Would it then make sense to remove the gender
> variable?
>
> And finally, is there a maximum of variables that can be used in a TBSS
> model?
>
>
> Thanks for your help,
>
> Niek
>
>
> ________________________________
> The materials in this message are private and may contain Protected
>Healthcare Information or other information of a sensitive nature. If you
>are not the intended recipient, be advised that any unauthorized use,
>disclosure, copying or the taking of any action in reliance on the
>contents of this information is strictly prohibited. If you have received
>this email in error, please immediately notify the sender via telephone
>or return mail.
---------------------------------------------------------------------------
---
De informatie opgenomen in dit bericht kan vertrouwelijk zijn en is
uitsluitend bestemd voor de geadresseerde. Indien u dit bericht onterecht
ontvangt, wordt u verzocht de inhoud niet te gebruiken en de afzender
direct
te informeren door het bericht te retourneren. Het Universitair Medisch
Centrum Utrecht is een publiekrechtelijke rechtspersoon in de zin van de
W.H.W.
(Wet Hoger Onderwijs en Wetenschappelijk Onderzoek) en staat geregistreerd
bij
de Kamer van Koophandel voor Midden-Nederland onder nr. 30244197.
Denk s.v.p aan het milieu voor u deze e-mail afdrukt.
---------------------------------------------------------------------------
---
This message may contain confidential information and is intended
exclusively
for the addressee. If you receive this message unintentionally, please do
not
use the contents but notify the sender immediately by return e-mail.
University
Medical Center Utrecht is a legal person by public law and is registered at
the Chamber of Commerce for Midden-Nederland under no. 30244197.
Please consider the environment before printing this e-mail.
________________________________
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
|