Just to be clear, if you created models that had only one or the other
of the 465.0102 or 465.0103 time points, and then compared them, they
definitely should be identical. However, I thought your question was
whether having both present impacted the design matrix (.mat file) in
which case you want to compare a model with just one of those time
points to your original model that includes both.
cheers,
-MH
On Thu, 2011-07-07 at 14:27 -0700, Katie Karlsgodt wrote:
> Hi,
> Well, I made each into it's own EV, loaded them up separately into GLM to make 2 separate models that each only included one of those files, and looked at the resulting .mat files. They are identical out to the 9th decimal point (which is as far as they go), so I'll just assume that this is below any meaningful threshold.
> thanks,
> Katie
>
>
> >>
> Eugene: Wouldn't this have something to do with the sub-sampling
> interval that I presume is lurking "under the hood" in constructing the
> EV in the digital domain? That is, there will be some interval that is
> too short for consecutive events to have an impact on the resulting EV,
> and I suspect that 0.0001 sec is indeed shorter than that interval.
> But, if the events were separated by say 0.1 sec, it might well be a
> different story.
>
> Katie: This seems easy enough to check empirically by removing the
> offending extra entry, saving the design, and then comparing the
> two .mat files to see if (and if so, how) they differ.
>
> cheers,
> -MH
>
> On Thu, 2011-07-07 at 19:25 +0000, Eugene Duff wrote:
> > Hi Katie -
> >
> >
> > I believe any timepoint that is assigned a height of 1 by any of the
> > individual events will be assigned to be one. i.e. the individual
> > events are not combined additively.
> >
> > Eugene
> >
> > --
> >
> > Eugene Duff, Phd
> > Analysis Group, Centre for Functional MRI of the Brain (FMRIB)
> > Nuffield Department of Clinical Neurosciences
> > John Radcliffe Hospital
> > University of Oxford, OX3 9DU
> >
> > Ph: +44 (0) 1865 222 523
> >
> > --
> >
> >
> >
> > On 7 July 2011 18:45, Katie Karlsgodt <
> [log in to unmask]
> > wrote:
> > Hi,
> > Through some kind of strange error, I have some 3 column
> > models that contain almost overlapping time points (the task
> > itself does not contain events this closely spaced). For
> > instance, there are 2 separate events that start at
> > 465.0102sec and 465.0103sec, respectively. The .0001
> > difference is far below anything I'd expect to be able to
> > measure, obviously, and I want to know what Feat does in this
> > situation- does this get modeled in some way as 2 events, or
> > is it basically equivalent to one event that is 4.0001s long
> > rather than 4s long (in which case I may not care as much
> > about the problem). I'm just curious how this works out. No
> > time points overlap between EVs, it is just near-duplicate
> > time points like the ones you can see below.
> >
> > thanks,
> > Katie
> >
> >
> > 11.5193 4 1
> > 38.5102 4 1
> > 66.0151 4 1
> > 76.0123 4 1
> > 81.5166 4 1
> > 92.5253 4 1
> > 104.5120 4 1
> > 118.5214 4 1
> > 129.0161 4 1
> > 135.0177 4 1
> > 151.0167 4 1
> > 155.5262 4 1
> > 182.0530 4 1
> > 193.0118 4 1
> > 202.0143 4 1
> > 220.5168 4 1
> > 226.0210 4 1
> > 240.0138 4 1
> > 254.0232 4 1
> > 261.5171 4 1
> > 283.0202 4 1
> > 292.0229 4 1
> > 316.5103 4 1
> > 324.0206 4 1
> > 324.0207 4 1
> > 336.5214 4 1
> > 390.5198 4 1
> > 412.0064 4 1
> > 418.5221 4 1
> > 429.5141 4 1
> > 443.5235 4 1
> > 453.0233 4 1
> > 465.0102 4 1
> > 465.0103 4 1
> > 473.5153 4 1
> >
> >
> > _________________________________________
> > Katherine H. Karlsgodt, Ph.D.
> > Center for Cognitive Neuroscience
> > Department of Psychiatry
> > University of California, Los Angeles
> >
> > email:
> [log in to unmask]
>
> > phone: 310-206-3019
> > fax: 310-794-9740
|