Hi Peter,
I'm afraid I'm still a little unclear what you are asking.
If you add a new EV into a model, but do not use it in the contrasts,
then it can have two effects in general. If it is very different
(uncorrelated) with the other EVs then all it can do is "soak up"
some of the otherwise unexplained signal in the data (i.e., reduces
the unmodelled noise, boosting the zstats for all contrasts in a
small or large way).
If it _is_ correlated with the other covariates then in general
adding the EV will reduce the zstats in your contrasts, as
statistical significance only derives from the part of the fitting
(of the other EVs) which is orthogonal to the new EV.
On the other hand, if you are wanting to include the new EV in
questions being asked by the contrasts, you will have to include it
in the appropriate contrasts.
Cheers, Steve.
On 20 Oct 2006, at 01:25, Peter Soros wrote:
> Hi Steve,
>
> sorry to bother you ... I guess you know what is wrong and have the
> right answer but I still can't get it working. I have attached the
> design matrix of a single run. It is a sparse sampling protocol, 51
> time points, 4 exp. conditions in blocks of 5 events each.
>
> I used "0" for EV 6 (here latency) in all 10 contrasts, but the
> model doesn't use EV 6. Do I have to change the value for EV 6? I
> played around and inserted 1 and -1, no change.
>
> All EVs are 3 column format, no convolution, no orthognalise, no
> added temporal derivative, no temporal filtering.
>
> Thank you so much for your help!!
>
> Peter
>
> <model.tiff>
>
>
> HI - I'm not sure _exactly_ what you're asking, but I think the
> answer is that you haven't changed the contrasts (or added new ones)
> to actually use the new EV?
> Cheers, Steve.
>
>
> On 18 Oct 2006, at 17:10, Peter Soros wrote:
>
>
>> Dear FSL experts,
>>
>> I would like to create an additional EV (response latency) for my
>> first-level FEAT analysis. I have an event-related sparse sampling
>> protocol (51 data points per run). I'd like to take the actual
>> latency
>> for each event and subject, rather than the mean latency in a group
>> analysis (as is shown in the FEAT manual).
>> How could that be done? I have created a text file in the custom
>> 3-column format. In the design matrix, the new EV shows up nicely,
>> the
>> original EVs, however, are not modified (and the FEAT maps are the
>> same as before).
>>
>> Thank you very much for your help,
>>
>> Peter
>>
>>
>
> __________________________________
>
> Dr. Peter Soros
> Postdoctoral Fellow
> Imaging Research
> Sunnybrook Health Sciences Centre
> 2075 Bayview Avenue, Room SG28
> Toronto, Ontario, M4N 3M5
> Canada
>
> tel 416.480.6100 ex. 3843
> fax 416.480.5714
>
> [log in to unmask]
>
> http://neuroactivity.org
>
>
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|