Thanks Dr. Harms. This is super helpful! On Tue, Mar 31, 2015 at 9:10 AM, Harms, Michael <[email protected]> wrote:
> > The AR(1) correction is a scaling applied during the r-to-z > transformation > > This is from nets_netmats.m > > RtoZcorrection=1/std(grotZ); > netmats=0.5*log((1+netmats)./(1-netmats))*RtoZcorrection; > > So, if you knew the value of RtoZcorrection you could convert the HCP > distributed netmats back to r-value. BUT, that value wasn't saved. > > If you want the netmats expressed as correlations (r-values) why not > just run: > > ts_dir='PATH_TO_TIMESERIES'; > nRuns = 4; > ts = nets_load(ts_dir,0.72,1,nRuns); > netmat1 = nets_netmats(ts,0,'corr'); %% Second argument 0, to leave > netmats as r > > It takes a little bit of time, but not *that* much, and you only need to > do it once. If I recall correctly, when I did it on the full set of > time-series from the HCP 500 release, it took something like an hour or two > total for the nets_load and nets_netmats steps. > > cheers, > -MH > > -- > Michael Harms, Ph.D. > ----------------------------------------------------------- > Conte Center for the Neuroscience of Mental Disorders > Washington University School of Medicine > Department of Psychiatry, Box 8134 > 660 South Euclid Ave. Tel: 314-747-6173 > St. Louis, MO 63110 Email: [email protected] > > From: Yizhou Ma <[email protected]> > Date: Monday, March 30, 2015 6:54 PM > To: "Harms, Michael" <[email protected]> > Cc: Stephen Smith <[email protected]>, "[email protected]" > <[email protected]> > > Subject: Re: [HCP-Users] netmats in HCP > > Dear Dr. Harms, > > Thank you very much for your codes. They are very helpful for me to > understand how FSLNets works. I am able to reproduce individual netmats > that are perfectly correlated with HCP distributed netmats. > > I also tried the mean netmat of full correlations netmats from the 4 > runs. They also give me netmats that are perfectly correlated with HCP > distributed netmats (by perfectly correlated I mean Pearson's > correlation=1.0000 in matlab). Thus the AR(1) process does not seem to > affect the result very much. > > Since I still do not understand the scaling of the z values in HCP > distributed netmats, may I know how I can transform them back to r values, > so that to get an idea of the size of correlation? (to run nets_netmats > with z=0 (leave netmats as r)) is too time consuming for this purpose). > > Thanks, > Cherry > > On Mon, Mar 30, 2015 at 1:34 PM, Harms, Michael <[email protected]> wrote: > >> >> Hi, >> You'll want to refer to the relevant functions in the FSLnets package. >> >> I did something very similar to what you are trying to do recently. >> Here is the code that you want to use if you are trying to replicate the >> distributed netmats (pconn's) from the distributed time-series: >> >> ts_dir='PATH_TO_TIMESERIES'; >> nRuns = 4; >> ts = nets_load(ts_dir,0.72,1,nRuns); >> netmat1 = nets_netmats(ts,1,'corr'); >> >> Of note, the estimation of the AR(1) correction factor that is used >> internally in nets_netmats as part of the r-to-z transformation is the >> median over all subjects, so if you are trying to duplicate the HCP >> distributed netmats "exactly", you correspondingly need to load the full >> set of distributed time-series as part of your 'nets_load' command. Even >> then, there will be a very small residual difference in the resulting >> output from nets_netmats due to the random seed component of the AR(1) >> simulation (i.e., running nets_netmats twice does not itself generate >> identical results when using its r-to-z conversion option). >> >> cheers, >> -MH >> >> -- >> Michael Harms, Ph.D. >> ----------------------------------------------------------- >> Conte Center for the Neuroscience of Mental Disorders >> Washington University School of Medicine >> Department of Psychiatry, Box 8134 >> 660 South Euclid Ave. Tel: 314-747-6173 >> St. Louis, MO 63110 Email: [email protected] >> >> From: Yizhou Ma <[email protected]> >> Date: Monday, March 30, 2015 12:26 PM >> To: Stephen Smith <[email protected]> >> >> Cc: "[email protected]" <[email protected]> >> Subject: Re: [HCP-Users] netmats in HCP >> >> Thank you all. I don't see a comment on correction for temporal >> autocorrelation in the PTN document though. Could someone please point me >> to a page where this is explained? >> >> Thanks, >> Cherry >> >> On Mon, Mar 30, 2015 at 12:18 PM, Stephen Smith <[email protected]> >> wrote: >> >>> yup - that also affects the scaling - see FSLNets doc and the PTN >>> release doc. >>> Cheers. >>> >>> >>> On 30 Mar 2015, at 18:17, Glasser, Matthew <[email protected]> >>> wrote: >>> >>> Hi Steve, >>> >>> What about the correction for temporal autocorrelation? >>> >>> Matt. >>> >>> From: Stephen Smith <[email protected]> >>> Date: Monday, March 30, 2015 at 12:15 PM >>> To: Yizhou Ma <[email protected]> >>> Cc: "[email protected]" <[email protected]> >>> Subject: Re: [HCP-Users] netmats in HCP >>> >>> Hi - one thing is that we estimate (z versions of) netmats separately >>> for each 15min run and then average the 4 netmats to give a single netmat >>> per subject. >>> Cheers. >>> >>> >>> On 30 Mar 2015, at 18:14, Yizhou Ma <[email protected]> wrote: >>> >>> Hi Timothy, >>> >>> Thank you for pointing that out. I typed that wrong but I used the >>> correct transformation in matlab. As I said my transformed z scores are >>> actually almost perfectly linearly correlated with HCP netmats, though the >>> latter is much larger. I want to understand why the latter is larger and >>> why the two are not exactly correlated. >>> >>> Thanks, >>> Cherry >>> >>> On Mon, Mar 30, 2015 at 12:08 PM, Timothy Coalson <[email protected]> >>> wrote: >>> >>>> A quick possibility: if you have pasted in the formula you used, I see >>>> an order of operations problem: .5*ln(1+r)-ln(1-r) means >>>> (.5*ln(1+r))-ln(1-r), >>>> where the usual formula is .5*ln((1+r)/(1-r)), which after some log >>>> identities becomes .5*(ln(1+r)-ln(1-r)). >>>> >>>> Tim >>>> >>>> >>>> On Mon, Mar 30, 2015 at 11:02 AM, Yizhou Ma <[email protected]> wrote: >>>> >>>>> Dear HCP experts, >>>>> >>>>> I am trying to reproduce the individual netmats from HCP500-PTN so >>>>> that I am sure where the numbers come from. I used individual node >>>>> timeseries in /ts2/subjID and did correlation in matlab. I then used >>>>> fisher's z transformation: .5*ln(1+r)-ln(1-r). The resulting netmat is >>>>> different from what is provided in *_netmat1/. However, they almost have a >>>>> linear relationship. It seems to me that HCP is not using the same z >>>>> transformation I have used. The transformation seems more like >>>>> 7*ln(1+r)-ln(1-r). >>>>> >>>>> I have downloaded FSLNets but could not identify which function was >>>>> used to generate individual netmats in the first place. The example script >>>>> seems to be about group-level netmats only. >>>>> >>>>> Could you please share with me how exactly the numbers in individual >>>>> netmats were generated? >>>>> >>>>> Thanks, >>>>> Cherry >>>>> _______________________________________________ >>>>> HCP-Users mailing list >>>>> [email protected] >>>>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >>>>> >>>> >>>> >>> _______________________________________________ >>> HCP-Users mailing list >>> [email protected] >>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >>> >>> >>> >>> >>> --------------------------------------------------------------------------- >>> 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) >>> [email protected] http://www.fmrib.ox.ac.uk/~steve >>> >>> --------------------------------------------------------------------------- >>> >>> Stop the cultural destruction of Tibet <http://smithinks.net/> >>> >>> >>> >>> >>> _______________________________________________ >>> HCP-Users mailing list >>> [email protected] >>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >>> >>> >>> ------------------------------ >>> The materials in this message are private and may contain Protected >>> Healthcare Information or other information of a sensitive nature. If you >>> are not the intended recipient, be advised that any unauthorized use, >>> disclosure, copying or the taking of any action in reliance on the contents >>> of this information is strictly prohibited. If you have received this email >>> in error, please immediately notify the sender via telephone or return mail. >>> >>> >>> >>> >>> --------------------------------------------------------------------------- >>> 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) >>> [email protected] http://www.fmrib.ox.ac.uk/~steve >>> >>> --------------------------------------------------------------------------- >>> >>> Stop the cultural destruction of Tibet <http://smithinks.net> >>> >>> >>> >>> >>> >> _______________________________________________ >> HCP-Users mailing list >> [email protected] >> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >> >> >> ------------------------------ >> >> The materials in this message are private and may contain Protected >> Healthcare Information or other information of a sensitive nature. If you >> are not the intended recipient, be advised that any unauthorized use, >> disclosure, copying or the taking of any action in reliance on the contents >> of this information is strictly prohibited. If you have received this email >> in error, please immediately notify the sender via telephone or return mail. >> > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
