Back to the main page.
Bug 2447 - plotting tutorial lacks a test function
| Status | CLOSED FIXED |
| Reported | 2014-01-22 09:51:00 +0100 |
| Modified | 2018-03-10 21:35:25 +0100 |
| Product: | FieldTrip |
| Component: | documentation |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Windows |
| Importance: | P5 normal |
| Assigned to: | |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | |
| See also: |
Jörn M. Horschig - 2014-01-22 09:51:45 +0100
anything to be done or can it just be removed? bug 185 seems to not help much here see http://fieldtrip.fcdonders.nl/tutorial/plotting
Jan-Mathijs Schoffelen - 2014-01-25 09:12:43 +0100
Do you know whether there's already a test_tutorial_plotting? If not, I suggest to make it and quickly check whether the two matlab-code lines, pertaining to the fixme, are really needed. If not, I suggest to remove.
Jan-Mathijs Schoffelen - 2014-01-25 09:31:58 +0100
I believe Arjen commented in #105, Arjen would you mind looking into this?
Arjen Stolk - 2014-01-25 11:16:25 +0100
(In reply to Jan-Mathijs Schoffelen from comment #2) Seems like a fossil. I have just removed it.
Jan-Mathijs Schoffelen - 2014-01-25 17:43:58 +0100
have you checked whether the test_tutorial_plotting exists?
Jan-Mathijs Schoffelen - 2014-02-20 15:42:31 +0100
Arjen seems to have diagnosed the problems, but failed to report the outcome of the check of the existence of the plotting tutorial. test_tutorial plotting does not seem to exist. It would make sense to make it
Jan-Mathijs Schoffelen - 2014-03-06 15:21:07 +0100
it would be useful to have it, because it tests in a single shot a variety of plotting functions. Nietzsche, would you mind having a go at it. No high priority though.
- 2014-03-06 15:26:07 +0100
okay.
- 2014-03-19 10:19:39 +0100
will also update the tutorial in certain places to make it clear which plot function produces which figure (plot)
- 2014-03-19 14:46:48 +0100
As discussed in FT meeting: (1) update plots - singleplot TFR (2nd figure) - beamforming plots - use those from beamforming tutorial - ERF cluster plot - use new plots (2) for cluster plot code just load the stat structure and provide plotting code. There's no need to start from preprocessing because we now have tutorials (this code was very old) (3)remove code that removes .dof (4)TFRFIC/TFRFC is now known as freqFIC/freqFC (variable names changed at some point but this tutorial wasn't updated) (5)colorbar issue is to do with cfg.render (check FAQ) (6) dccnfilename will be placed under utilities -i can then use this function (7) implement some short code for creating plots when using beamforming techniques. -> all variables required for this tutorial can go in the plotting folder (already created)
- 2014-03-19 16:00:00 +0100
As discussed in FT meeting: (1) update plots = - singleplot TFR (2nd figure) - FIXED - beamforming plots - use those from beamforming tutorial - ERF cluster plot - use new plots - FIXED (2) for cluster plot code just load the stat structure and provide plotting code. There's no need to start from preprocessing because we now have tutorials (this code was very old) - erf: - FIXED - TFR: - problem (3)remove code that removes .dof - FIXED (4)TFRFIC/TFRFC is now known as freqFIC/freqFC (variable names changed at some point but this tutorial wasn't updated): found freqFIC_planar_cmb and freqFC_planar_cmb files under tutorial/cluster_permutation_freq. However, there's an error online 281 of ft_freqstatistics. > will look into this further. (5)colorbar issue is to do with cfg.render (check FAQ) - FIXED (6) dccnfilename will be placed under utilities -i can then use this function (7) implement some short code for creating plots when using beamforming techniques. only 2, 6, and 7 are left.
- 2014-04-09 14:45:06 +0200
Only issue 7 is left. The ortho plot and slice plot options from ft_sourceplot work fine, but I have problems with creating the surface plot from the beamforming data.
Arjen Stolk - 2014-04-09 15:04:22 +0200
Hi Nietsche, Did you try surfaceplotting with an up-to-date version of FT? There was an issue that has been solved on monday. Yours, Arjen
- 2014-04-09 20:43:02 +0200
@ Arjan, thanks for the tip. I did a new svn update, and played around with cfg.surfdownsample but I think my matlab session was just not running smoothly anymore so I just started a new session and all was fine. Now I know how to make use of cfg.surfdownsample :) @JM/Arjan: ft_clusterplot checks the input for cfg.correcttail and cfg.alpha. However, it assumes that the user will always specify these two fields, which may not be the case e.g., user just wants to look at clusters without correcting. To deal with this I added an
Jan-Mathijs Schoffelen - 2014-04-16 13:51:58 +0200
Cfg.alpha as an option is a bit unlucky in this context. Probably we need to implement rather a cfg.highlightprobthreshold (as a vector, just like the other highlightxxxx options). Related to this, we discussed that as such the interpretation in clusters does not make sense to begin with. although we don't want to disable the function we should build in a warning note in the tutorial about this, and probably make a faq: can I interpret shape and extent of a cluster obtained with a cluster-based permutation statistical approach? Answer: NO. Next, check whether there is sufficient documentation of the correcttail option.
- 2014-10-31 00:52:00 +0100
The original task of plottin function (items 1-7) was completed and added to svn a while ago. The change on line 140 (comment 13) was removed. However, bug isn't closed because of two new tasks 1. add a warning in tutorial that we shouldn't interpret the data in terms of clusters. 2. write an FAQ on why not, and link that to the tutorial 3. check documentation on correcttail - Done. There is sufficient documentation here: http://fieldtrip.fcdonders.nl/faq/why_should_i_use_the_cfg.correcttail_option_when_using_statistics_montecarlo Wil get cracking on 1 and 2.
- 2014-11-12 17:15:17 +0100
FAQ on "how not to interpret results from a cluster-based permutation test"
- 2014-11-12 17:18:52 +0100
Here is what I've written so for the FAQ http://fieldtrip.fcdonders.nl/faq/how_to_not_interpret_results_from_a_clusterbasedpermutation_test @ Robert - could you please look through this ?
Robert Oostenveld - 2014-11-13 16:59:41 +0100
(In reply to n.lam from comment #17) done. I will follow it up on the fieldtrip-bugs email list
- 2014-11-14 09:44:07 +0100
Following Robert's editing of the FAQ, along with other contributions, I believe this FAQ is looking really good. Robert also tagged the FAQ into all statistics tutorials. All assigned tasks in this this bug have been completed. Closing.
Jörn M. Horschig - 2014-11-14 10:31:24 +0100
top!
Jörn M. Horschig - 2014-11-19 11:04:45 +0100
can we extend the answer by the question: what does the p-value of my clusters tell me then?
Robert Oostenveld - 2014-11-19 11:59:28 +0100
(In reply to Jörn M. Horschig from comment #21) good question. However, the p-value is not "of the cluster", but represents the probability of observing the data given H0. Admitted: confusing is that we do assign them (in the output data structure) to clusters and to channel-time-frequency points (or voxels). So your question could also be asked about the p-value on the voxels. I suggest we follow up in a FT meeting.