Print

Print
Hi Anderson,

Just to update in case anyone has a similar issue, the data did save when I used the gamma approximation option. I thought I had tried this before as someone else had a similar issue that was resolved in this way, but apparently not!

Out of interest, and I understand this would be undermining the whole point of the approximation in the first place, but if I use the gamma approximation with a large number or permutations would this be effectively the same as running without the approximation? Meaning that the gamma approximation wouldn’t be adding much because there is a large amount of permutations to base it on? 

All the best,
Sam

 

On 12 Jul 2019, at 14:14, Samuel Berry <[log in to unmask]> wrote:

Dear Anderson,

thats okay, thank you for taking a look. I tried on a different computer with more up-to-date software but still no joy strangely. I am trying with the cifti files instead (probably a better idea anyway). Thats a kind offer, if this doesn’t work either and I really have exhausted all options I may take you up on that, although this complete fail would not be favourite solution either. Many thanks for your time.

Best wishes,
Sam

On 8 Jul 2019, at 21:31, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Sam,

I don't know what may be going on... are you using recent versions of Octave (or Matlab) and PALM? If so, and the problem keeps ongoing, can you try in a different computer? If nothing else, I can exceptionally offer to run the analysis for you... send me an email off-list and we can find a solution... (though surely that wouldn't my favorite one).

All the best,

Anderson


On Fri, 5 Jul 2019 at 09:48, Samuel Berry <[log in to unmask]> wrote:
Hi Anderson,

many thanks for this. Does this mean that it got to the end and saved it successfully as well? That's great but also frustrating if so, as I am still having problems when PALM is computing the p-values. The error is not coming directly from PALM as such, but it is from a function that is called by PALM. The error is generated by the ‘griddedInterpolant' function, which is called by ‘interp1', which is called by 'palm_pareto>andersondarling’. The line in PALM with the issue is:

 i2 = interp1(i1,ptable,A2,’linear’,’extrap’).

I have created a breakpoint in matlab at this point and I can see that A2 has the value ‘inf’. A2 I see is created earlier using the variable Z. Z in this case is a 1x13 double and the first value is 0 (all other values are non-zero, e.g 0.1164). i1 has 8 non-zero values (eg. 0.3464).

Does this help at all in diagnosing the problem? I am trying to work backwards from the error to understand what might be happening but I am finding it quite confusing. I am happy to send over a config file or whatever other files you might need if this helps. I guess if its working for you then this could be a problem with my version of matlab perhaps? Once again, thank you for your time it is very much appreciated. 

Best wishes,
Sam.

On 27 Jun 2019, at 15:16, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Sam,

It worked for me with up to 1000 permutations.

Also, this message, regarding coordinates, don't seem to be coming from PALM... maybe there's another program running that has been wrapped around it?

All the best,

Anderson


On Mon, 24 Jun 2019 at 05:42, Samuel Berry <[log in to unmask]> wrote:
Hi Anderson,

thank you for doing this its much appreciated. It also runs okay for me if I use 10 permutations, the problem is when I use more (say 100) the error I get is: 

’The coordinates of the input points must be finite values: Inf and NaN are not permitted’

I think this might be due to the fact I get extremely low p-values. I have tried to remove the seed-region, which was very highly correlated with itself, but I am still having the same problem. Do you have any recommendations for how to proceed? 

Best wishes,
Sam

On 23 Jun 2019, at 21:45, Anderson M. Winkler <[log in to unmask]> wrote:

Hi Sam,

Thanks for sending. I wasn't able to replicate, though. It run fine for me -- that is, a one-sample t-test in a file that has 1071 maps. I run with 10 permutations using Octave 4.4.1 and the code run through completion with no error... unsure what may be going on... perhaps if you try in a different computer that has different software versions/configurations?

All the best,

Anderson


On Wed, 29 May 2019 at 08:27, Samuel Berry <[log in to unmask]> wrote:
Thank you Anderson, I have sent you an email with a link to share a dropbox folder. 

Best wishes,
Sam
From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Anderson M. Winkler <[log in to unmask]>
Sent: 14 May 2019 00:39
To: [log in to unmask]
Subject: Re: [FSL] PALM analysis errors with HCP data - data cannot be sorted
 
Hi Sam,

Would you mind sending your input files so I could have a look? You could use any file sharing service of your preference, just let me know the link then.

Thanks!

All the best,

Anderson

On Mon, 29 Apr 2019 at 10:33, Samuel Berry <[log in to unmask]> wrote:
Hello,

I have been having some issues when running one-sample PALM analysis on a 4d nifti image that contains 3T zmaps from a seed-based correlation analysis. The error I get is:

Error using palm_competitive
Data cannot be sorted. Check for NaNs that might be present, or precision issues that may cause over/underflow. If you are using “-approx tail”, consider adding “nouncorrected”.

Having read some other posts on this error I have taken the following steps.

1) Changed the seed to be ‘twist’
2) Run fslmaths -nan on my data before running the analysis
3) Converted the 4d file to ‘double’ using fslmaths <4dimg> <output> -odt double
4) Run with and without tail approximation, always with the -nouncorrected option.
5) Tried without the EB blocks
6) Used up to 10,000 permutations

Unfortunately I always get the same error. I went into the palm_competitive script and saved the variables created and it shows that I do indeed have NaNs (from the gg variable). I tried, just to see what happens, to bypass the error by replacing any NaNs with the mean of the ’S’ variable which did let me progress past the error. However my output is then a fwep image that is all ‘1’ and a tstat image that is just NaN. Do you have any idea what I am doing wrong? I have been stuck on this for a while now and can’t figure it out.

My PALM input is the following (with the variations I have mentioned above, and -n 5 being just for testing).

palm -i <4d_img>, -T, -eb <EB_file>,  -save1-p,  -n 5, -seed twist, -m <mask_file>, -o <output_file>

Many thanks,
Sam

########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1


To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1


To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1



To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1



To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1


To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1



To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1



To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1