 |
 |
Hello, Thanks for the advice. I use fslsplit to explode de original time series file into n files, each one corresponding to one volume. Then I use fslmerge to reconstruct time series files with different lengths (or durations), which then I input to FEAT to analyze. Yes, I pay attention to the nonlinearities. In fact, if there is some time tolerance, I already realized that is better to increase ISIs (instead of more events), to compel events separation. Kind regards, Jose Paulo Santos On Tue, 3 Jun 2008 08:31:28 +0100, Steve Smith <[log in to unmask]> wrote: >Hi, > >yes, unless FEAT complains about rank deficiency I would ignore the >eigenvalues and just use the "effect required" efficiency calculations. > >It is fine to reduce the number of timepoints in your data using >fslsplit (I'm not sure why you'd want to use fslmerge) to simulate a >shorted experiment and see what effect that makes on the efficiency >calculations. > >Note though, as described in the paper, the efficiency calculations >don't know anything about the nonlinearities that occur as the events >get closer and closer together, so be careful there. > >Cheers. > > >On 2 Jun 2008, at 15:44, Jose Paulo Santos wrote: > >> Hello, >> >> Sorry for the confusion. I use the term “efficacy” in the sense that >> I use >> these values to determine if a paradigm is efficacious or not, i.e., >> if >> they look OK and then I can pursuit with the paradigm, or not, and >> changes >> to the paradigm are required. >> >> It’s acceptable to use fslsplit and fslmerge to reduce the number of >> volumes in the data file to compare the resulting values of Effect >> Required and “Efficacy”? Or these comparisons aren’t possible? >> >> Kind regards, >> Jose Paulo Santos >> >> >> On Mon, 2 Jun 2008 07:27:16 +0100, Steve Smith <[log in to unmask]> >> wrote: >> >>> Hi - I'm still a little confused as we never call this "efficacy" - >>> this is a plot of the eigenvalues from a singular value decomposition >>> of the design matrix - it's not as directly useful or interpretable >>> as >>> the "Effect required" efficiency calculations - if they look ok then >>> that's good enough. >>> >>> Cheers. >>> >>> >>> On 1 Jun 2008, at 23:29, Jose Paulo Santos wrote: >>> >>>> Hello, >>>> >>>> Yes, I got these values from FEAT (please see attachment; this >>>> example is >>>> for 250 volumes). The Efficacy is the right most value in the >>>> diagonal (I >>>> didn’t consider the others) and the Mean Effect Required is the mean >>>> of >>>> Ci’s. >>>> >>>> Kind regards, >>>> Jose Paulo Santos >>>> >>>> >>>> On Fri, 30 May 2008 08:19:03 +0100, Steve Smith <[log in to unmask] >>>> > >>>> wrote: >>>> >>>>> Hi - I'm not sure where these Efficacy numbers come from - not from >>>>> the FEAT efficiency calculations I think? I'm not sure what these >>>>> are....... >>>>> Cheers. >>>>> >>>>> >>>>> >>>>> On 28 May 2008, at 19:10, Jose Paulo Santos wrote: >>>>> >>>>>> Hello, >>>>>> >>>>>> I'm trying to establish some kind of method to achieve to the >>>>>> optimal >>>>>> length in event-related paradigms. >>>>>> >>>>>> I designed a paradigm were I collected 500 volumes. The >>>>>> participant >>>>>> reported at the end that he was exhausted. So I know that 500 >>>>>> volumes were >>>>>> too much in the human perspective. By the end of the session, the >>>>>> probability of the participant still being concentrated in the >>>>>> task >>>>>> was >>>>>> quite low. >>>>>> >>>>>> With fslsplit and fslmerge commands, I reconstructed several >>>>>> files, >>>>>> until >>>>>> 250 volumes, and I analyzed each of them with Feat. The numerical >>>>>> data was: >>>>>> >>>>>> Volumes Efficacy Mean Effect Required >>>>>> 250 0.253 0.938 >>>>>> 312 0.241 0.847 >>>>>> 375 0.251 0.752 >>>>>> 438 0.245 0.740 >>>>>> 500 0.243 0.722 >>>>>> >>>>>> The rule of thumb for the Efficacy determines that it must be more >>>>>> than >>>>>> 0.200, and it was for all volumes. The rule of thumb for the >>>>>> Effect >>>>>> Required determines that it must be less than 0.800. This >>>>>> requirement is >>>>>> true only for 375 volumes or more. When this data is plotted, it >>>>>> seems >>>>>> that there aren’t significant increases in the Effect Required for >>>>>> more >>>>>> than 375 volumes. This is in line when the maps are analyzed: the >>>>>> 375 >>>>>> volumes map has activations more intense (just a bit) and more >>>>>> extended >>>>>> (just a bit) than the 438 and 500 maps. So, 375 volumes would be >>>>>> the >>>>>> optimal length (and duration) of the session. If there is some >>>>>> clearance, >>>>>> it would be better to increase the inter stimulus interval (null >>>>>> event) >>>>>> than to show more stimuli. >>>>>> >>>>>> Is this acceptable? Why the Efficacy didn’t change so much? >>>>>> Increasing >>>>>> ISIs would increase the Efficacy? >>>>>> >>>>>> Thanks in advance, >>>>>> Jose Paulo Santos >>>>>> >>>>> >>>>> >>>>> --------------------------------------------------------------------- -- >> --- >>>> - >>>>> 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 >>>>> --------------------------------------------------------------------- -- >> --- >>>> - >>>>> = >>>>> = >>>>> = >>>>> = >>>>> = >>>>> =================================================================== >>>> <design_cov.png> >>> >>> >>> ----------------------------------------------------------------------- --- >> - >>> 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 >>> ----------------------------------------------------------------------- --- >> - >>> = >>> = >>> = >>> ===================================================================== >> >> > > >-------------------------------------------------------------------------- - >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 >-------------------------------------------------------------------------- - >========================================================================