Activity for Joe Dien
-
Joe Dien posted a comment on discussion Help
Okay, I just sent you a link to a beta version with the fix.
-
Hi Shuyi, I'm not quite sure what you mean by "factor loadings for each participant for each component. "There is only one set of factor loadings for the entire dataset when you conduct a PCA. The Edit function provides the raw numbers if you use it to inspect a PCA result. Is that what you are asking for? As for the tsPCA.ept file, if you send me a copy of the file, I'd be happy to fix the problem.
-
Rank of data is zero? Something peculiar is going on, that's for sure! I'll need you to submit a full bug report so that I can look into it. At the start of the tutorial pdf is a section on bug reports that tells you what I'll need to troubleshoot.
-
I can see how the warning message could be ambiguous. I'll clarify it for the next release.
-
It doesn't sound like you followed the installation instructions. Please take another look at them. Cheers! Joe On 6/27/24 17:07, PJ Brancaleone wrote: I am attempting to launch the EP toolkit for the 1st time. I have downloaded the most recent version (EP_Toolkit 2993) and unzipped the contents into where I have my other EEG applications like eeglab as recommended in the tutorial pdf. Then, I opened Matlab and clicked "add to path" -> "Current folder", then tried to run the ep command. When I do,...
-
or to put it another way, it is telling you that you are supposed to use the "Add with Subfolders" button. Your not using it is in fact the problem. Be sure to read the instructions regarding EEGlab and FieldTrip as well.
-
It doesn't sound like you followed the installation instructions. Please take another look at them. Cheers! Joe
-
If you mean the values of the factor pattern and factor structure matrices, you can obtain them using the Edit function.
-
In the case of the tutorial, this was handled by choosing the "autoPCA" option. It chooses the peak time point as the time window, which in this example was "192-196 ms." The "176-224 ms" you are quoting was for the windowing of the averaged data, not the PCA data. That would be chosen in the normal manner. In this case, the rationale was 25 ms on either side of 200 ms, for a 50ms long window, which is reasonable for an N2. If you're wondering why the numbers aren't multiples of five, it is because...
-
Ah, that makes more sense! It depends on what kind of reliability you want. The EP Toolkit is already set up to provide SME and Dependibility measures for ERP measures. Truth to tell, I don't think either of these provide what you have in mind. They are both pretty specialized in their nature. PCA is layering another level of complexity, so I don't know of anyone who uses reliability measures with them, except for between session reliability (at the subject level). The EP Toolkit automatically provides...
-
Well, first I need you to be much more precise about how this dataset was generated. The devil is in the details, as they say. Please confirm that it was applied to individual trials, and then if so, describe how this was conducted. You're not describing a conventional EP Toolkit analysis, so there are many possibilities here.
-
and which part of the statistics are you trying to do the split-half reliability on? The PCA solution itself or the ANOVAs?
-
Hi, sure! I'll need more information to fully respond though. So first of all, was your PCA applied to the subject averages or to the individual trials? And for the split-half reliability, do you indeed need the reliability at the trial level? I ask, because computing the PCA at the subject level, but then trying to do trial-level things with the average-level results would be tricky. Joe
-
No problem. Once the temporal PCA has been performed, it no longer matters what time points are chosen for the window. All time points reflect the same factor scores. You'll get the exact same p-values no matter what time points you choose. Please see my 2004 and 2012 papers for more explanation of the math behind this process. Joe
-
Thanks for your kind words! I'm afraid I don't really understand your question. You set a window in the Windowing function. You enter in the start and end ms of the window you want.
-
oops! oversight in the code generating cell means and standard errors for the output file. I'm sending you a fix.
-
Ah got it! Sorry, when I get a figure with things circled, I'm used to it being the "after" errors being highlighted, not the "before" good values. This was a bug due to the subject specs not being handled right when there are multiple sessions per subject. I just sent you the fix. Give it a try and let me know if you run into any further issues!
-
Okay, well my initial inspection shows no difference between the subject specs in the .ept average file and in the .ept PCA file. So what is the problem? There's no difference before and after the autoPCA as far as I can tell.
-
Hi, in order to troubleshoot, I need the before as well as the after, so to speak. The way it works is I run it through the same steps as you did and see at what point it goes awry. I can't do this if I don't have a copy of the average file that you performed the PCA upon. See the instructions in the Tutorial on how to perform a bug report.
-
Hi, sorry for the delay! I was out sick. I can't access your files because I don't have an account on your University's system. You need to set the file so that it is public access.
-
That'd be a "depends" I'm afraid. So for example, a channel interpolation is only as good as the surrounding channels. Channels on the edge of the montage are more likely to have issues because of the missing information about what is outside of the montage. Choosing only channels in common avoids the interpolation issues but will rob the PCA of much of its power as temporal PCA relies heavily on the differences in topography across the channels to distinguish between ERP components, so the less...
-
Yes, definitely standardize the recordings first. You can use the Edit function to do both. The Samples pane allows you to resample as desired and the Channel Coordinates button of the Overview pane allows you to interpolate the channels from one montage to another. Then you'll be ready to go!
-
Okay, I think I've fixed everything and have sent you the link to a fixed version. I'll include the changes in the next EP Toolkit release. It should now provide better support for .fif files and EGI Hydrocel-33 channel systems.
-
No problem! It was a genuine bug, if a bit of a specialized scenario. Fixed now. The ANOVA bug seems to come up because I broke the between-group code for more than one factor when I added the support for conventional ANOVAs. I'm sending you a fix for you to try out.
-
I still can't replicate how your file runs into the triggering conditions of having no eloc info but I'm sending you a possible fix. Try it out and let me know.
-
Okay, I'm starting to narrow in on what is going on with your PCA file. But there is something that still doesn't add up. The error is happening because, according to your test file, the PCA file didn't have the eloc electrode coordinates info but did have the name of the CED file and when it tried to get the info from it, it got thrown because that bit of code wasn't expecting to be doing this for a PCA file, which normally inherits the eloc information from the original average file. What is confusing...
-
Happy to look at it! I'll need a copy of the file, a screenshot of your settings, the output of "ver", and the startup text after you type in ep. See tutorial section on bug reports.
-
With the files you sent me, I can see that the crash is happening because the code has somehow gotten to a spot that it should not have reached and on my computer is working fine. If you are willing to invest some more time in figuring this out, could you put a breakpoint on line 4452 of the ep_readData file (click on the line number and a red dot will appear) and then try reading the file again? When it stops (because it has reached the breakpoint), save the contents of the workspace again like...
-
I just released 2.96, which includes a fix for the unexpectedly large History records that you ran into.
-
I assume you mean "PCA" not "ACP"? Anyway, I see the crash in your video but I still can't replicate it so I can't fix it. Could you send me the output of "ver"? Also the messages you get in the command window when you start up the EP Toolkit. Also, in the tutorial troubleshooting section, it explains how to use Stop on Errors to send me a test file containing the state of the variables at the moment of the crash. That would be extremely helpful.
-
Glad to hear it fixed the problem! I wasn't able to replicate your error with the new file. You'll need to give me more information. Were you just reading it using the Read function? Could you send a screencap? As for the size, I used the new Memory pane of the Edit function and it revealed that the reason it is so large is because of the new History logging. It looks like you made a bunch of edits to the file (adding specs information) and it is logging each time you added a single value and it...
-
Okay, I figured it out! It was a bug introduced by my effort to support ANOVA text files without headers. Download the fixed beta from the link I gave you and try it out.
-
I have indeed increased the number of bootstrapping runs. For some analyses, I found it was needed to maintain the desired level of stability. Needless to say, faster computers helpful! I'll also need a screenshot of your ANOVA function settings.
-
Yep that's right! Use the Edit function to look at them. There is a session field that keeps track of which session they come from.
-
Oops! That was part of the backward compatibility code for the History field. Try downloading it again from the link I gave you. It should be fixed now. Yes please do report the ANOVA problem if you're still having it with the beta build. This is the time to do it!
-
You're in luck! I was just getting the next release ready so I took a moment to add the edit information to the Trials subpane of the Edit function. I'm sending you the beta to try out.
-
Simplest is to use the View > Scan function and look at the Bad Trial checkbox.
-
Okay, fixed! It'll be in the EP Toolkit beta I e-mailed you. If you downloaded it prior to getting this message, then download it again. Joe
-
I tried reading it and didn't have any problems. It may be an issue I've already fixed. I'm going to send you a copy of my current beta so you can see if it fixes your problem. I'm getting ready to post a new release in a few days.
-
No problem! I'll need a copy of the file. Joe
-
I'm sorry, I don't understand the question. What would be the reason for needing to reamp via the Edit function instead of just selecting the montage in the dropdown menu? You've explained why you don't want to do the former but not why the latter doesn't work for you.
-
Sure! Just use the Subject pane of the Edit function. See the tutorial for more information.
-
I'm sorry, I only support Matlab 2018b and onwards. FYI, the problem is that 2015 doesn't support string type variables.
-
I tried to send you the link to the beta but your sourceforge e-mail address doesn't work. Send me an e-mail address and I'll send you the link. Joe
-
I've been thinking about this very topic! We use the error bars for two reasons. The first is to get a sense for which conditions are significantly different. For repeated measures, we would just use the Loftus-Masson number as all the cells have the same standard error for this purpose (since they all share the same error term) and so this number is already provided. However, we might also want to provide bars to indicate the individual variability of each cell. In order to support this purpose,...
-
Hi Chao, my general outlook is that it's not really possible to say that one approach is the best with all data. Even a sub-optimal approach may be particularly well suited to isolate a particular ERP component due to its unique characteristics. So I do recommend average reference for PCA (Dien, 2012) but mostly because temporal (not spatial) PCA essentially does it anyway and also because one of my primary uses for PCA is source analysis and there again the average reference is required. If one...
-
The problem is that you need to update to the current version (2.93). It fixes a bug in that function. Always make sure to be using the current version. Joe
-
Yes, that is exactly right re windowing. You should indeed get the same results. How did you go about getting the factor scores? And what version of EP Toolkit are you using? More information required to answer question.
-
Hi, I've fixed the bug. It was due to a change in how grand average data are represented in the EP Toolkit and will send you the fix. As for the questions, you cannot pull out the individual subjects after having used a grand average file. That would be equivalent to your telling me that the mean reaction time of a dataset is 500 ms and then asking me to pull out the individual reaction times from that one single number. In any case, my 2010 paper showed you don't get as good results from a grand...
-
Your Matlab is too old. You should be getting a message about it when you start EP Toolkit and requirements are also in the tutorial. You need 2018b onwards. Joe
-
Good to hear! I'll try putting in a fix for this as well. I should have the new release out soon.
-
Someone else had the same problem and I just made a fix. I'll send it to you so you can try it out. Let me know if it works. I plan to have another release soon.
-
I fixed the error you ran into with the Edit function. I wasn't able to track down the problem with the jack-knife as you only sent me one subject average file. I need you to send me the actual spatial PCA file in .ept format that you are trying to jack-knife so I can reproduce your error exactly. You can use the Save function to create the .ept file.
-
I need to work out a better way of doing this. Could you send me a sample EEG file to go with the ced file you sent me for testing purposes?
-
Including Cz is fine. The reference channel is just another EEG channel so really I should have said 129 channels. Pooling the three frontal channels is a a little more problematic if you are double-counting some channels by including them in this extra 130th channel as well. Anyway, main problem is that my code won't recognize this as being EGI data if you have 130 channels rather than 129. My suggestion is to use the Edit function to reassign this 130th channel to being of type REG. This will inform...
-
Sorry about the problem! The current EP Toolkit enhances electrode positioning by co-registering them with a canonical set of electrode positions via 10-5 names. The EGI system is an exception to this positioning system so my software has special procedures for handling them. They are recognized by EGI's standard practice of naming them E1-E128. Also, EGI mff files come with three fiducial points (nasion and two mastoids). I'm guessing that the pipeline through BVA discarded the fiducials and renamed...
-
Complicated questions! The initial temporal PCA is separating EEG activity based on timecourse, so anything that has the same timecourse will end up in the same temporal factor, even if they are entirely different processes. When you say that the first temporal factor is significant, what you really mean is that it was significant at the channel with the largest amplitude, which is what the EP Toolkit automatically finds for you. That doesn't mean that the activity captured by the first temporal...
-
Okay, the next step is to read the instructions in the tutorial file on how to report a bug to me. It lists some information I need.
-
Okay, fixed! I'm sending you a link to the current beta with the fix separately. Joe
-
Just for the sake of others reading this thread, it looks like the website isn't showing his entire message for some reason. Here is the complete set of errors he was getting: Warning: It appears that the subdirectories of fieldtrip were not added to the path using ft_defaults. This means that there are subdirectories are on the path that should not be, which can destabilize Matlab. See tutorial file for details. Warning: You have a binary ICA program installed. The file is called: ica_linux There...
-
The error messages just mean what they say. You need to follow the installation instructions in the tutorial file. It's in the Documentation folder. Joe
-
It's not finding a file it needs that is in the EEGlab folder. When you start up EP Toolkit, what error messages are you getting?
-
HI Harry, this is quite a bit more than I need. All I need is the PCA file that you're trying to look at with the Factors tab. I just need to replicate your error message. Could you send me just that one .ept file? Sorry for the trouble! Joe
-
Yes, that's exactly right. When we apply a temporal PCA, for example, we're separating out the portion of the timepoints that are acting the same. That's the definition of a factor. By that same token, if you apply inferential statistics to any time point of a factor, they will be the same. Or mathematically, what you're doing is applying statistics to the same set of factor scores, scaled at each time point by the factor loading for that time point. The scaling by the factor loading does not affect...
-
Sorry about that! Send me test data and I'll fix it.
-
No problem! Although my name is Dien. :) You have two options. one is to set up the data in an EEGlab .study structure, which is what it uses for multi-subject/multi-condition datasets. Then it can be read directly into EP Toolkit. The other option is to set up individual files (using whatever file format) for each subject/condition and then use the Single File option in EP Toolkit to read them in.
-
Oh, one other thing is that EOG channels are just EEG channels that are positioned around the eyes so normally they are treated just like any other EEG channel and should be retained in the PCA. The artifact correction routines take care of the eye artifacts (or at least the EP Toolkit one does). If you have some reason to drop them (like the hardware of some systems reference EOG channels differently than the other channels) then yes drop them by all means.
-
If you see any problems with how the EP Toolkit read in your data, let me know and I'll fix it. Cheers! Joe
-
The EP Toolkit should be reading in all artifact marking information. I didn't follow all of that, but the best thing is always to check yourself as you know your own data best. Use the Edit function to see what is going on with the channels and so forth. Use the View>Scan function to see how bad channels and trials have been marked. PCA will only include channels that are EEG and it will automatically exclude channels and trials that are marked bad. If you have EEG data in there that you don't want...
-
It would take a bit of doing to do what you're asking for. One option would be to just redo the spectral analyses inside of the EP Toolkit. Another thought is that if you just click on the dB button the View pane, you can change the units for visualization. Just choose "cm" rather than "dB" and it won't be doing any transformations to your data. Ditto for analyses via the Window function.
-
Sorry for delay! I'm getting a new release ready. I should have a fix out to you tomorrow. Joe
-
Hi Javier, the problem is that the channels are the rows but you've got EP Toolkit thinking the channels are the columns. You need to go into the Preferences and change the setting. That's why you have the error message saying it thinks you have 300 channels. See tutorial for more information. I'll update the error messages to make them more helpful for this situation.
-
Sorry about the problems! The issue is that you lost the EPdataset file that keeps track of everything in the EPwork directory. Fortunately you can salvage the files. Just move them out of EPwork, replace the .mat ending with .ept, and you're good to go! You can read them back into the EP Toolkit.
-
Glad to hear you were able to solve it! Let me know if there is anything further I can help with. Joe
-
I still can't replicate your report. I'm not doubting that you are running into something, just that this seems to be particular to your system. On the theory that maybe existing binaries are exempted when upgrading to Catalina, I downloaded a new copy of FieldTrip and did the ft_defaults procedure but still no problem. Then I tested the possibility that it being on Dropbox made a difference but that wasn't it either. I looked on the web more and some folks are saying that when Catalina downloads...
-
I bit the bullet and updated my Mac to Catalina (I'm mostly using a Linux computer these days). I wasn't able to replicate the problems you ran into. On the Security & Privacy pane of the system Settings, on the General subpane, I have the "App Store and identified developers" checked. Apparently that is enough for my system. Is that what you have? You mentioned something about "allowed each of them" so it sounds like you're doing something different? I don't see any place to do anything like that...
-
Hi, I've been avoiding updating to Catalina because of these sorts of issues. This is not just FieldTrip as ep_WJGLMml.mexmaci64 is one of my EP Toolkit files. The problem is that the Mac OS has become extremely tight about security. Mex files are Matlab binary files that have been compiled for extra speed (I got something like a ten times boost by doing this but it was a lot of work as I had to recode everything in C++). On the FieldTrip mailing list, I found the following posting by Sendy Caffarra:...
-
If you read the EP Toolkit tutorial, the bug report section tells you how to take screenshots. In this case all I really need to know is whether you were indicating that this was an average file or a single-trial file. I was able to proceed anyway and fix the problem. I'm e-mailing you the fix.
-
Hi Harry, sorry for these problems. The maxColors error is a bug that was reported to me right after I posted the release. It only affects new installs. I'm going to have to add that case to my testing script. Anyway, I reposted 2.89 (as 2.891) right afterwards so if you redownload the EP Toolkit, it'll be solved. As for the problem with the EEG files, I'd have to click on each and every file individually to download all of these. Could you please package them all into a single .zip file? If you...
-
Hi Harry, no prob! This is a very complex file format. In order to troubleshoot I'll need a copy of your file and a screenshot of your EP Toolkit settings and the ver output. If you're talking about the .study file you provided earlier, you also need to provide the .set files that go along with it. Basically you need to send me everything that EEGlab needs to load the data.
-
Sorry about that! I'm sending you a link to a fix. I'll be posting an update in the next day or so. Joe
-
I took a look and it seemed to work fine. Please give more information. When you say "not applying" what exactly are you doing and what exactly are you seeing on the screen? For example, what you said could mean "I changed the number but it didn't remember the change when I checked it again" or it could mean "when I ran autoPCA it reported using the original .005 number" or it could mean "when I ran autoPCA it reported using the .001 number but didn't actually use that as the threshold" or any number...
-
error had nothing to do with the ced. I'm sending you a fix.
-
see the electrodes directory for examples Sent from my iPhone On Jan 24, 2020, at 15:59, Emily Perkins eperkins@users.sourceforge.net wrote: Hi again, sorry to ask a million questions and thank you again for your help thus far. I'm still erroring out on the loading-in-ced-file step, and I think it's probably because of the reference. For one dataset I'm working with, I average-referenced, and for the other, it's mastoid-referenced. For both datasets, everything is already processed and I'm inputting...
-
no problem! A .ced file really is just a text file with a .ced at the end of the name rather than a .txt. EEGlab is pretty picky about that but I suppose I could modify EP Toolkit so it doesn't care about the suffix. Anyway, just rename it. You may need to set the OS to Show the Suffix as sometimes they are set to be invisible, making the suffixes even more mysterious. As for the VEOG and HEOG, you can set them to BAD. See the tutorial. If they are referenced to the the same reference as the rest...
-
Sorry about this! There were some bugs that got by my testing. I'll be posting a bugfix update later today I think. I'm sending you my current build. Let me know if you run into any more issues!
-
Sorry for the problems! It's difficult to accommodate all the possible variations in text files. In this case the data were space delimited with an initial space that caused the code to think there was an extra column. I'm sending you a new version that should fix it. Also a fix for the three-step PCA scree. Note that even I haven't actually used the three-step option so you're heading into new territory there.
-
turned out I hadn't tested the single file mode enough. Also crashing in PCA. Made a bunch of fixes. I'm sending you fixes. Once they work, I'll post a new release.
-
found a case where single file mode can crash in the read function. Also added output to the command window to help confirm settings. Am sending link. Let me know if you run into any further issues!
-
that's exactly right! you'll also need to rename the file names so that the letter positions are consistent across files. Sorry for all the frustration!
-
So In Sub001error01.txt the characters with the frequency are 11-12 and in Sub002correct01.txt they are 14-15. So what did you type into the fields? or are these not even the files you used?
-
I'm sorry, I can't help until you send me what I've asked for with exactly what you are doing. Just now you sent me one thing but when I tested it and told you the problem, you tell me you did something different. Please send me the test files, a screenshot of exactly what your settings are, and the output of the ver command. For example, looking at the names of the files you sent me, it is impossible for you to be setting the Single File Mode fields correctly. But I can't tell if this is the problem...
-
Just looking at the screenshot, it appears you didn't set the subject and cell fields for the single file mode controls.
-
also, in order to help with bug reports, I will always need the information from the "ver" command and a screenshot of your settings. See the tutorial for what I need to troubleshoot and how to get it for me. As an example of why I will always need test files, it is possible is has nothing to do with your data or with the Toolkit. It could be a problem with how the text editor generated the .txt files. Matlab has trouble handling unicode text files for example. I can't diagnose this kind of problem...
-
I'll modify the labeling so it's clearer.
-
Hi, it's not a bug. Those are two different retention criteria. See tutorial.
-
Sure! whatever you tried to do. I just need a couple subjects and conditions and frequencies so eight files would suffice but more wouldn't hurt. Please do this ASAP. It's holding up the release...
-
Your timing is good. I'm in the middle of ironing the bugs out of a new release. Could you send me some test files?
-
No current guidance on this. For now it's just meant to provide a way to flag you to problems with a specific subject session. Also it's a way to judge differences in noise level between systems and so forth.
-
never mind, I got it fixed! I'll send it to you tomorrow.
1 >