Hi JLS
Either heavy temporal filtering or the temporal derivative might be
fitting to some noise or signal spikes at the onset/offset of the
block. The temporal derivative regressor fits independently of the
main regressor, so can sometimes effect fits other than the temporal
shifts that it is primarily intended to account for.
Eugene
On 25 October 2010 21:05, Jeremy Smith <[log in to unmask]> wrote:
> Not sure if this list accepts attachments, so please let me know if I need to do it another way.
>
> The green trace in the [hopefully-]attached screenshot should be a unitary "lump," rather than two cusps and a "valley."
>
> Regards,
> JLS
>
>
>
>> Hi - I've never seen this and can't think what this might be - can you
>> send in an example?
>> Cheers.
>>
>>
>> On 22 Oct 2010, at 17:17, Jeremy Smith wrote:
>>
>> > I searched the FSL Archives for this, but came up empty-handed.
>> >
>> > We're convolving a simple boxcar vector with a gamma function to
>> create an EV model in first-level FEAT. In the FEAT report, however,
>> the convolved waveform appears to "clip" at the top of the y-axis, such
>> that the peak and a few time points around it get inverted... so
>> instead of a nice "hill" we get two cusps and a "valley." Is this
>> merely a graphing issue, or will it affect our analysis? What can be
>> done about it?
>
|