Hi Michael, This has worked wonderfully. Thank you all for your very helpful messages. Xavier. ________________________________ From: Harms, Michael [mha...@wustl.edu] Sent: Tuesday, January 31, 2017 11:01 AM To: Xavier Guell Paradis; hcp-users@humanconnectome.org Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance?
Yes. -- 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: mha...@wustl.edu From: <hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>> on behalf of Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Tuesday, January 31, 2017 at 9:59 AM To: Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Michael, Thank you so much for all your help. I have seen that for each task, each subject has multiple cope1 files (under different cope folders; for example for the EMOTION task subject 100206 has the following folders, each with a cope1.dtseries.nii file: cope1.feat, cope2.feat, cope3.feat, cope4.feat, cope5.feat, cope6.feat). The Contrasts.txt file for EMOTION shows the following: FACES, SHAPES, FACES-SHAPES, neg_FACES, neg_SHAPES, SHAPES-FACES. If I am interested in FACES-SHAPES, should I use the cope1.dtseries.nii file that is inside the cope3.feat folder (given that FACES-SHAPES is the third contrast listed in the Contrasts.txt file)? Thank you, Xavier. ________________________________ From: Harms, Michael [mha...@wustl.edu<mailto:mha...@wustl.edu>] Sent: Monday, January 30, 2017 9:11 PM To: Xavier Guell Paradis; Glasser, Matthew; Elam, Jennifer; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Cc: Burgess, Gregory Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? No, we were proposing this: M = cifti map of mean(of Level 2 individual subject copes) S = cifti map of std(of Level 2 individual subject copes) Cohen’s d = M/S 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Monday, January 30, 2017 at 6:41 PM To: "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>>, Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? I think this would be done from a level3 file; maybe I'm wrong. If I understand this correctly, the three steps below would give a Cohen's d map. Have I understood it right?: 1st) take this file: HCP_S900_787_tfMRI_ALLTASKS_level3_zstat1_hp200_s2_MSMAll.dscalar.nii 2nd) transform this file into a cope1 file (Michael said he may be able to make this file available; "I can make the equivalent “cope1” file from the Level 3 ‘flameo’ available via Box") 3rd) in the cope1 file, do (x-mean)/SD for every data point Thank you very much, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Monday, January 30, 2017 7:26 PM To: Xavier Guell Paradis; Harms, Michael; Elam, Jennifer; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Cc: Burgess, Gregory Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? I assumed you were talking about level3 files. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Monday, January 30, 2017 at 6:25 PM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "Harms, Michael" <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Hi Matthew, I am sorry, I didn't fully understand your previous message. Do you mean that the two steps that I mentioned in my last message are correct? Thank you very much, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Monday, January 30, 2017 7:22 PM To: Xavier Guell Paradis; Harms, Michael; Elam, Jennifer; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Cc: Burgess, Gregory Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Those are the level 2 copes. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Monday, January 30, 2017 at 6:20 PM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "Harms, Michael" <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Matthew, Thank you very much for the suggestion. To make sure that I understand this correctly; would this be the correct analysis?: 1st) Obtain a group cope1 file of the S900 group task contrasts (in a previous message, Michael said he could make this available from the S900 group task contrasts zstat maps: "I can make the equivalent “cope1” file from the Level 3 ‘flameo’ available via Box" 2nd) For each data point of the group cope1 file, calculate (x-mean)/SD. This gives a Cohen's d map. Is this correct? Thank you, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Monday, January 30, 2017 6:51 PM To: Xavier Guell Paradis; Harms, Michael; Elam, Jennifer; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Cc: Burgess, Gregory Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? No because the non-optimal scaling will divide out in the mean(cope)/std(cope) ratio. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Monday, January 30, 2017 at 4:35 PM To: "Harms, Michael" <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>>, Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Michael, Wouldn't using the cope files have the problem that "the released versions are not optimally scaled (because of a non-optimal intensity bias field correction)" (as written by Matthew before in this conversation)? Or would this not matter if Cohen's d were calculated from cope1 files? Thanks, Xavier. ________________________________ From: Harms, Michael [mha...@wustl.edu<mailto:mha...@wustl.edu>] Sent: Monday, January 30, 2017 12:30 PM To: Elam, Jennifer; Glasser, Matthew; Xavier Guell Paradis; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Cc: Burgess, Gregory Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Hi, Just wanted to mention that I’m not promoting the computation of Cohen’s d effect size maps by dividing the z-stat maps by sqrt(N) as a “formal” solution. Since the z-stat are computed using ‘flameo’ and multi-level variance modeling, I think the “proper” way to compute Cohen’s d effect size maps would be from first principles — i.e., the mean (across subjects) divided by the std (across subjects), of the Level 2 copes. And even that might have some issues, due to the family structure (resulting in a underestimate of the std across subjects). We’ll give this some thought, and aim to include Cohen’s d effect size maps of the task contrasts as part of the group average maps for the S1200 release. 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>> Date: Monday, January 30, 2017 at 11:21 AM To: Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>>, Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Posting this off-list thread about computing effect size task maps for the S900 from the group average zstat maps back to the list in case looking through it is of interest to other users. If the discussion continues, please include the list address in the responses so others can follow. Thanks, Jenn Jennifer Elam, Ph.D. Scientific Outreach, Human Connectome Project Washington University School of Medicine Department of Neuroscience, Box 8108 660 South Euclid Avenue St. Louis, MO 63110 314-362-9387<tel:314-362-9387> e...@wustl.edu<mailto:e...@wustl.edu> www.humanconnectome.org<http://www.humanconnectome.org/> ________________________________ From: Harms, Michael Sent: Sunday, January 29, 2017 10:39 PM To: Glasser, Matthew; Xavier Guell Paradis Cc: Burgess, Gregory; Elam, Jennifer Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Or perhaps, rather than showing p=0.05 as a contour, show the Cohen’s d = XX contour, to get away from the problem where, with 787 subjects, huge chunks of cortex will have p<0.05, even if the Cohen’s d effect size is tiny in many places. 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>> Date: Saturday, January 28, 2017 at 1:50 PM To: Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Significance is a quantification of “likelihood of truth” whereas effect size is a quantification of “importance.” With large numbers of subjects, the uncertainty of the measure diminishes and thus it can be considered “true” in the absence of confounds, but that does not say whether we should care about it or not. As Mike says, it is unclear that there is a threshold of effect size that is anything other than arbitrary (just as with significance we by convention set an arbitrary threshold of p=0.05) and application dependent. With a properly normalized beta map, you can see which areas are strongly different from each other (and therefore have strong gradients between them), however even gradient strength is a continuous number. This is one of the reasons that I don’t advocate thresholding these maps, though it may be helpful to indicate the p=0.05 level as a contour on top of the map. Peace, Matt. From: "Harms, Michael" <mha...@wustl.edu<mailto:mha...@wustl.edu>> Date: Friday, January 27, 2017 at 10:26 PM To: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>>, Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? As an approximation, you could divide the z-stat by sqrt(787). https://www.researchgate.net/post/Is_cohen_d_equal_to_z_statistics_How_can_I_calculate_cohen_d_using_Z_scores But the z-stat you have is based on ‘flameo’ and multi-level variance modeling, so I not sure exactly how much the traditional statistical relationship between z and d applies. I still don’t have a sense of how you are going to *operationally* define whether you “trust the functional topography”. The effect size is going to be continually varying, just like the z-stats. So, are you going to “trust” a cohen’s d effect size of 0.3, but not trust one of 0.25? Any threshold you set is arbitrary. 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Friday, January 27, 2017 at 5:01 PM To: Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Hi Michael, Thanks for your reply. I will try to calculate a group average Cohen's d effect size map for the task contrasts. Do you know if it is possible to do this by using only the group average task contrasts zstat map (i.e. this file: HCP_S900_787_tfMRI_ALLTASKS_level3_zstat1_hp200_s2_MSMAll.dscalar.nii), or do you think it will be necessary to go back to the data of each individual subject? Thank you very much, Xavier. PS: What I meant by "trust the functional topography" is the following. Using a very high z threshold (e.g., 14) in the group average task contrast z maps shows some areas of activation for each task. I was wondering whether this finding alone can be understood as a "true" observation (i.e. whether it is true that the areas that are most activated in task X are areas A and B, and that the areas that are most activated in task Y are areas C and D), and Matt very helpfully explained that a measure of effect size would be necessary to know that activation in a particular area is not only statistically significant but also "important to that task". As Matt wrote, "Regions with a large effect size in a task are likely important to that task (and will probably also meet criteria for statistical significance given a reasonable amount of data)." As such, I understood that an effect size map would allow me to trust the functional topography observed after using a very high z threshold. I would trust the zmap with a threshold of z>14 for a particular task contrast if the effect size map with a given threshold also revealed the same areas of activation. ________________________________ From: Harms, Michael [mha...@wustl.edu<mailto:mha...@wustl.edu>] Sent: Friday, January 27, 2017 5:36 PM To: Xavier Guell Paradis; Glasser, Matthew Cc: Burgess, Gregory; Elam, Jennifer Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? How do you operationally define “trust the functional topography”? I think a Cohen’s d effect size map is an attractive thing, but I don’t have that precomputed at this time. 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Friday, January 27, 2017 at 3:42 PM To: Michael Harms <mha...@wustl.edu<mailto:mha...@wustl.edu>>, "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>> Cc: "Burgess, Gregory" <gburg...@wustl.edu<mailto:gburg...@wustl.edu>>, "Elam, Jennifer" <e...@wustl.edu<mailto:e...@wustl.edu>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Hi Michael, Thank you very much for your reply. This is the zstat file I am using: HCP_S900_787_tfMRI_ALLTASKS_level3_zstat1_hp200_s2_MSMAll.dscalar.nii The reason I would like to have a group average effect size map is to know whether I can trust the functional topography seen when using a very high z threshold in these group average task contrast zmaps (for example, when using a z threshold of 14). As Matt very helpfully explained, the large number of subjects in S900 zmaps results in very large z values. Therefore, the way to know whether we can trust these very high z values (>14) is not tests of statistical significance (given that z values >14 will most likely be statistically significant even after correcting for multiple comparisons and family structure) but measures of effect size. My plan was to make sure that the regions that show the highest z values are also regions that show large enough effect sizes. I can see from your comment on "non optimal scaling" that a group average cope1 file might actually not be useful for this purpose. I would be extremely grateful if you could suggest an alternative approach I could try in order to calculate a "correct" group average effect size map for the task contrasts. I would also be very interested in hearing your thoughts regarding alternative ways of knowing whether one can trust the functional topography that can be seen when using a very high z threshold in these group average task contrast zmaps. Thank you very much for your help, I really appreciate it. Xavier. ________________________________ From: Harms, Michael [mha...@wustl.edu<mailto:mha...@wustl.edu>] Sent: Friday, January 27, 2017 4:09 PM To: Glasser, Matthew; Xavier Guell Paradis Cc: Burgess, Gregory; Elam, Jennifer Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Hi Xavier, Which specific zstat file from the group average package are you using? I can make the equivalent “cope1” file from the Level 3 ‘flameo’ available via Box. Not quite sure what you’ll do with it however. In light of the “non optimal scaling” that Matt mentioned, you can’t really convert it to % BOLD change (at least not in a manner that is meaningfully consistent across grayordinates). So I’m curious, what do you intend to do with this map once you have it? For the group level maps in the final 1200 release, I intend to make a Cohen’s d effect size map, which will be self-normalizing spatially. 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: mha...@wustl.edu<mailto:mha...@wustl.edu> From: <hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>> on behalf of "Glasser, Matthew" <glass...@wustl.edu<mailto:glass...@wustl.edu>> Date: Friday, January 27, 2017 at 2:26 PM To: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? We need the folks who ran the analysis and did the packaging to help out here. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Friday, January 27, 2017 at 2:24 PM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Thank you again for the reply. Is there a way to access data that was produced but not packaged up? Thank you for your help, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Friday, January 27, 2017 3:21 PM To: Xavier Guell Paradis; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? It was produced, whether or not it was packaged up is a separate matter. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Friday, January 27, 2017 at 2:18 PM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Matt, Thanks for the reply. Did you mean that the pipelines produce group average effect size map and that these must be somewhere as part of the HCP public data; or were you referring to the individual subject effect size maps? Thanks, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Friday, January 27, 2017 3:01 PM To: Xavier Guell Paradis; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? I assume these are available somewhere because the pipelines produce them, but I didn’t make these group average results. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Friday, January 27, 2017 at 10:04 AM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Matt, Thank you again for your reply. I have been able to find cope1 files for single subject task contrasts (e.g. cope1 file for working memory contrasts of subject 996782), but not for the S900 group (e.g. I have not been able to find a cope1 file for the S900 group for working memory contrasts). I was wondering: a) Is there any task contrast effect size map available for the S900 group? (even if they are not optimally scaled) b) If not, would it be possible to generate a task contrast effect size map by using available S900 group data (e.g. the task contrasts zstat maps of the S900 group), or would it be necessary to go back to the data of each individual subject? c) If it is necessary to go back to the data of each individual subject, which approach would you suggest to combine all cope1 files of each subject of the S900 group into one effect size map of all subjects? Would something like normalizing the cope1 file of each subject (using wb_command as written below) and then averaging all normalized cope1 files work? Or would something as simple as averaging all cope1 files work? wb_command -cifti-reduce <input> MEAN mean.dtseries.nii wb_command -cifti-reduce <input> STDEV stdev.dtseries.nii wb_command -cifti-math '(x - mean) / stdev' <output> -fixnan 0 -var x <input> -var mean mean.dtseries.nii -select 1 1 -repeat -var stdev stdev.dtseries.nii -select 1 1 -repeat Thank you very much, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Thursday, January 26, 2017 6:53 PM To: Xavier Guell Paradis; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? The files called cope1 or beta are an effect size measure, however the released versions are not optimally scaled (because of a non-optimal intensity bias field correction). We plan to correct this in the future. Peace, Matt. From: Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Thursday, January 26, 2017 at 5:41 PM To: Matt Glasser <glass...@wustl.edu<mailto:glass...@wustl.edu>>, "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: RE: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear Matt, Thank you very much for your very helpful reply. I will have to investigate this topic more, but is there any approach you would suggest to obtain effect size maps from the S900 group HCP data? I was wondering if the zstat data of the S900 group task contrasts could be converted to effect size values without having to go back to the individual subjects. Thank you very much, Xavier. ________________________________ From: Glasser, Matthew [glass...@wustl.edu<mailto:glass...@wustl.edu>] Sent: Thursday, January 26, 2017 5:33 PM To: Xavier Guell Paradis; hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Subject: Re: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Standard error scales with sample size, standard deviation does not. Things like Z, t, and p all also scale with sample size and are measures of statistical significance via various transformations. Thus for a large group of subjects, Z and t will be very high and p will be very low. Z, t and p are thus all not biologically interpretable, as their values also depend on the amount and quality of the data. In the limit with infinite amounts of data, the entire brain will be significant for any task, but wether a region is statistically significant tells us little about its importance functionally. Measures like appropriately scaled GLM regression betas, %BOLD change, or Cohen’s d are biologically interpretable measures of effect size because their values should not change as sample size and data amount go up (rather the uncertainty on their estimates goes down). Regions with a large effect size in a task are likely important to that task (and will probably also meet criteria for statistical significance given a reasonable amount of data). A common problem in neuroimaging studies is showing thresholded statistical significance maps rather than effect size maps (ideally unthresholded with an indication of which portions meet tests of statistical significance), and in general focusing on statistically significant blobs rather than the effect size in identifiable brain areas (which should often show stepwise changes in activity at their borders). Peace, Matt. From: <hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>> on behalf of Xavier Guell Paradis <xavie...@mit.edu<mailto:xavie...@mit.edu>> Date: Thursday, January 26, 2017 at 3:46 PM To: "hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>" <hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org>> Subject: [HCP-Users] Very large z values for task contrasts in S900_ALLTASKS_level3_zstat file: what does this mean in terms of statistical significance? Dear HCP team, I have seen that the zstat values for tasks contrasts are very large in the HCP_S900_787_tfMRI_ALLTASKS_level3_zstat1_hp200_s2_MSMAll.dscalar.nii file, to the point that one can observe areas of activation in task contrasts by setting very high z value thresholds (e.g., a z threshold of +14). I think (please correct me if I'm wrong) that the z values of the S900 file are very large because the group is very large, therefore the standard deviation is very small (given that there will be less variability in a group if one takes a very large group of people rather than a small group of people), and if the standard deviation is very small then even small differences from the mean will lead to very large z values. I was wondering what implication does this have in terms of statistical significance. A z value of 14 or larger would correspond to an extremely small p value, i.e. it would be extremely unlikely to observe by chance a measure which is 14 times the standard deviation away from the mean. Would it therefore be correct to assume that the areas that we can observe in the S900 tfMRI_ALLTASKS task contrasts with a very high zstat threshold (e.g., 14) are statistically significant, without having to worry about multiple comparisons or family structure? Thank you very much, Xavier. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org<mailto:HCP-Users@humanconnectome.org> 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. ________________________________ 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. ________________________________ 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 HCP-Users@humanconnectome.org<mailto:HCP-Users@humanconnectome.org> 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. ________________________________ 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. ________________________________ 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. ________________________________ 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 HCP-Users@humanconnectome.org<mailto:HCP-Users@humanconnectome.org> 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. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
