 |
 |
Hi, Yes, the sigmasquared estimate has already had the whole model removed. Ideally you should take into account the pre-whitening, but this will be a reasonable approximation. All the best, Mark On 24 Jun 2009, at 18:19, Luis Hernandez wrote: > thanks, steve, that might do the job ... but won't that also > include the variances from the other regressors? or is sigmasqareds > the variance after removing the whole model ( that is , var(Y - > X*beta) ) > > -L > > > On Jun 23, 2009, at 2:29 PM, Steve Smith wrote: > >> Hi - you can just take the residuals (take root of sigmasquareds) >> and divide by sqrt(ntimepoints) to get standard deviation of the >> mean_func - I'm pretty sure that's close enough....? >> >> Cheers. >> >> >> On 23 Jun 2009, at 10:50, Luis Hernandez wrote: >> >>> right ... there is not Beta_hat_0 in FSL - it's just my way of >>> saying estimate of baseline regressor parameter - (the one I'm >>> after). In case you're wondering, I'm trying to use it to >>> estimate spin density. >>> >>> thanks again >>> >>> -L >>> >>> >>> On Jun 23, 2009, at 12:23 PM, Mark Jenkinson wrote: >>> >>>> Hi, >>>> >>>> In FSL there is no pe0 or beta_hat_0 file as we remove >>>> the mean prior to setting up the GLM. Hence there is >>>> no constant (mean) regressor and so no parameter estimate >>>> (or variance) associated with that. >>>> >>>> I'm not sure where you are getting Beta_hat_0 from. >>>> >>>> All the best, >>>> Mark >>>> >>>> >>>> On 23 Jun 2009, at 16:28, Luis Hernandez wrote: >>>> >>>>> thanks, Mark. >>>>> >>>>> Beta_hat_0 is the same as the mean (assuming no pre-whitening), >>>>> but I don't think the variance estimate of Beta_hat_0 the same >>>>> as the variance of the signal. oh, well... >>>>> >>>>> >>>>> -Luis >>>>> >>>>> >>>>> >>>>> >>>>> On Jun 23, 2009, at 10:59 AM, Mark Jenkinson wrote: >>>>> >>>>>> Hi, >>>>>> >>>>>> If you are talking about the mean signal as the baseline >>>>>> regressor then >>>>>> I'm afraid the answer is no. We do not remove the mean signal >>>>>> via a >>>>>> regressor, but instead remove it from the signal and the model >>>>>> in the >>>>>> pre-statistical filtering. Hence we do not generate values or >>>>>> variances >>>>>> for it. But it is easy to calculate the mean using fslmaths >>>>>> and -Tmean >>>>>> on the original data if that is all you want. >>>>>> >>>>>> All the best, >>>>>> Mark >>>>>> >>>>>> >>>>>> On 23 Jun 2009, at 15:49, Luis Hernandez wrote: >>>>>> >>>>>>> Dear FSL experts, >>>>>>> >>>>>>> Is there a way to get FSL to dump the parameter estimate for >>>>>>> the baseline regressor into a file, like pe0? (It looks like >>>>>>> the mean_func file might do the trick ... is this right?) BUT >>>>>>> more importantly, how about its variance estimate? can that be >>>>>>> dumped out? FEAT writes out varcope files based on the >>>>>>> specified contrasts but I don't see how to make a contrast >>>>>>> that uses the baseline regressor, since it is kind of "hidden" ? >>>>>>> >>>>>>> I apologize if I missed something obvious. thanks in advance >>>>>>> >>>>>>> -Luis >>>>>>> >>>>>>> -- >>>>>>> ( http://www.eecs.umich.edu/~hernan ) >>>>>>> >>>>>> >>>>>> >>>>> >>>>> -- >>>>> ( http://www.eecs.umich.edu/~hernan ) >>>>> >>>> >>>> >>> >>> -- >>> ( http://www.eecs.umich.edu/~hernan ) >>> >> >> >> --------------------------------------------------------------------------- >> 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 >> --------------------------------------------------------------------------- >> >> > > -- > ( http://www.eecs.umich.edu/~hernan ) >