Hi Estephan,
What are the names of the files that you’re looking into? The fwep files contain p-values, and with the (recommended) -logp option, these are always positive.
All the best,
Anderson
--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
hcp-users+...@humanconnectome.org.
To view this discussion on the web visit
https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/96806701-c5bb-46ed-913d-acf8ee405631n%40humanconnectome.org.
A p-value is a measure of statistical significance not of effect size. A p-value tells you only the likelihood that a test is not due to random variation, not how strong an effect is or in what direction. Despite what you might find in the literature, showing maps of thresholded p-values is not best practice.
Matt.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/757f4625-8688-4aed-bd1f-390e2a9327d5n%40humanconnectome.org.
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.
What files are available to you?
Matt.
I guess none of those are really effect size measures (as the plain tstat files will still be scaled by the square root of sample size). Ideally, PALM would also output an effect size measure, but the tstat files would at least give you positives and negatives.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/0b60ca3f-95dc-46e3-8912-1b5d499cb3fbn%40humanconnectome.org.
If you use the PALM version from Github (do a "git clone https://github.com/andersonwinkler/PALM), the option -saveglm has been updated to output Cohen's d, which is a standardized measure of effect size.
The -log(p) has a linear-ish relationship with the effect size (at least more linear than the p-values in their original range 0-1). This means that the colorscale isn't too different between -log(p) and Cohen's d or t-stat, so for viewing it's a quick option.
Maybe see if the new version of PALM has similar results?
Matt.
From: Estephan Moana <esteph...@gmail.com>
Date: Wednesday, January 27, 2021 at 11:55 AM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: "Glasser, Matthew" <glas...@wustl.edu>, "Winkler, Anderson (NIH/NIMH) [E]" <anderson...@nih.gov>, Estephan Moana <esteph...@gmail.com>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
I see. I did look into generating effect size maps for these group means by following the methods in this reference: Geuter, S., et al. (2018). "Effect Size and Power in fMRI Group Analysis." bioRxiv: 295048. It basically consists of generating CIFTI files with the mean and the standard deviation for each group using the concatenated CIFTI COPE files from 2nd level (the input file for the PALM analysis), then calculating Cohen's d CIFTI files for each group by calculating the mean/std dev. Is this a reasonable approach in your opinion?
I did find that the Cohen's d effect sizes were rather large, and also an interesting finding. One of the groups had effect sizes ranging from -0.84 to 1.48, however it did not any activation surviving thresholding (all -logp values were less than 1.301). Please see below the screenshots showing the Cohen's d map for this group, the overlay and map settings and the fwep results in -logp format. What are your thoughts on this?



On Wednesday, January 27, 2021 at 11:27:11 AM UTC-6 glas...@wustl.edu wrote:
I guess none of those are really effect size measures (as the plain tstat files will still be scaled by the square root of sample size). Ideally, PALM would also output an effect size measure, but the tstat files would at least give you positives and negatives.
Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Wednesday, January 27, 2021 at 11:22 AM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: "Glasser, Matthew" <glas...@wustl.edu>, "Winkler, Anderson (NIH/NIMH) [E]" <anderson...@nih.gov>, Estephan Moana <esteph...@gmail.com>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
The ones generated by the PALM call I posted previously:
- F-test: fstat, fstat_uncp, fstat_fdrp, fstat_fwep
- Group mean contrasts (c1-3): tstat, tstat_uncp, tstat_fdrp, tstat_fwep
I attached an screenshoot of the folder with the files listed. Thank you.
On Wednesday, January 27, 2021 at 11:18:03 AM UTC-6 glas...@wustl.edu wrote:
What files are available to you?
Matt.
From: Estephan Moana <esteph...@gmail.com>
Date: Wednesday, January 27, 2021 at 11:15 AM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: "Glasser, Matthew" <glas...@wustl.edu>, "Winkler, Anderson (NIH/NIMH) [E]" <anderson...@nih.gov>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
That's a good point, Matt. So if I want to look into the group mean activations/deactivations, how would you suggest me to proceed?
On Wednesday, January 27, 2021 at 11:11:57 AM UTC-6 glas...@wustl.edu wrote:
A p-value is a measure of statistical significance not of effect size. A p-value tells you only the likelihood that a test is not due to random variation, not how strong an effect is or in what direction. Despite what you might find in the literature, showing maps of thresholded p-values is not best practice.
Matt.
From: Estephan Moana <esteph...@gmail.com>
Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Wednesday, January 27, 2021 at 11:09 AM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: "Winkler, Anderson (NIH/NIMH) [E]" <anderson...@nih.gov>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
Thanks for you reply, Anderson. These are the files I am looking at:
- results_parcellated_Glasser2016-MMP1.0-ModifiedS1subregions_s2_dat_tstat_fwep_c1.pscalar.nii
- results_parcellated_Glasser2016-MMP1.0-ModifiedS1subregions_s2_dat_tstat_fwep_c2.pscalar.nii
- results_parcellated_Glasser2016-MMP1.0-ModifiedS1subregions_s2_dat_tstat_fwep_c3.pscalar.nii
Since fwep files always contain positive p-values when using the -logp option, how should I interpret these p-values with regards mean group activations/deactivations?
Thank you.
Estephan
On Wednesday, January 27, 2021 at 10:48:18 AM UTC-6 Winkler, Anderson (NIH/NIMH) [E] wrote:
Hi Estephan,
What are the names of the files that you’re looking into? The fwep files contain p-values, and with the (recommended) -logp option, these are always positive.
All the best,
Anderson
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/96806701-c5bb-46ed-913d-acf8ee405631n%40humanconnectome.org.
--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcp-users+...@humanconnectome.org.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/757f4625-8688-4aed-bd1f-390e2a9327d5n%40humanconnectome.org.
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.
--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcp-users+...@humanconnectome.org.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/0b60ca3f-95dc-46e3-8912-1b5d499cb3fbn%40humanconnectome.org.
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.
Either the particular calculation you used for Cohen’s d isn’t correct, or your N is rather small.
Cheers,
-MH
--
Michael Harms, Ph.D.
-----------------------------------------------------------
Associate Professor of Psychiatry
Washington University School of Medicine
Department of Psychiatry, Box 8134
660 South Euclid Ave. Tel: 314-747-6173
St. Louis, MO 63110 Email: mha...@wustl.edu
From: "Glasser, Matthew" <glas...@wustl.edu>
Reply-To: "hcp-...@humanconnectome.org" <hcp-...@humanconnectome.org>
Date: Wednesday, January 27, 2021 at 12:04 PM
To: Estephan Moana <esteph...@gmail.com>, HCP-Users <hcp-...@humanconnectome.org>
Cc: "Winkler, Anderson (NIH/NIMH) [E]" <anderson...@nih.gov>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
|
* External Email - Caution * |
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/30693993-AC2C-4CB7-BD34-3656AC18AC50%40wustl.edu.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/cd3f4031-689d-4789-aa55-01d71f1f4069n%40humanconnectome.org.
Hi Reza,
Thanks for commenting but this isn't correct really. PALM doesn't convert negative t-values to zero (or to anything). They are reported as usual, along with their p-values.
Perhaps you are alluding to two-sided p-values, i.e, instead of testing if t >= t_critical, one would be interested if |t| >= t_critical. That can be accomplished with the option -twotail, and even here, the negative t-stats are still reported as they come.
Having said that, yes, doing two contrasts (e.g., A>B and also A<B), then correcting using -corrcon, is recommended, but not mandatory.
Happy to add more information if anything isn't clear.
Thanks again!
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/D12FE8B1-E952-403F-B8B1-4E6B0194E4C8%40nih.gov.
Hi Reza,
Depends on the data. If the data are all positive or mostly positive, it's possible that the means are then all positive, hence all positive t-stats.
Error! Filename not specified.
Error! Filename not specified.
Error! Filename not specified.
To view this discussion on the web visit
https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/CAPiNLe7tL8BdTS30QzMtGfe5ySwkoLTvxyS_%3DnBhyG5vaTB7iA%40mail.gmail.com.
Hi Lisanne,
Which version of Matlab/Octave are you using? Please, try replacing that offending line (line 42 in palm_cohen.m) for:
cattmp = vertcat(M*plm.eC{y}{m}{c}{o},0)
mintmp = min(cattmp,1)
maxtmp = max(cattmp,1)
Xr = maxtmp - mintmp;
I believe it will work (though I’m not testing at this time).
All the best,
Anderson
From:
'Lisanne Jenkins' via HCP-Users <hcp-...@humanconnectome.org>
Date: Sunday, August 4, 2024 at 7:48 PM
To: HCP-Users <hcp-...@humanconnectome.org>
Cc: Winkler, Anderson (NIH/NIMH) [E] <anderson...@nih.gov>
Subject: Re: [hcp-users] fMRI parcellated group analysis PALM results values range
External Mail
This email originated outside of The University of Texas Rio Grande Valley.
Please exercise caution when clicking on links or opening attachments.
To view this discussion on the web visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/8555e9fc-13c5-4a39-b21a-7c9d9cbcc8a0n%40humanconnectome.org.
Hi,
Thank you for your reply. I worked out the issue – I had to install some matlab add-ons.
Best,
Lisanne
--
You received this message because you are subscribed to a topic in the Google Groups "HCP-Users" group.
To unsubscribe from this topic, visit
https://groups.google.com/a/humanconnectome.org/d/topic/hcp-users/bwmocGR8YB8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to
hcp-users+...@humanconnectome.org.
To view this discussion on the web visit
https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/DS0PR05MB97419B59923741711A6E564E82BE2%40DS0PR05MB9741.namprd05.prod.outlook.com.