Activity for Dominik Bach
-
Dominik Bach posted a comment on discussion Help
Hi Mark. PsPM has moved to github: https://github.com/bachlab/PsPM Kindly post your follow-up questions there and please make sure you use the latest version of the software. Regarding your question, I was not sure what you wanted to infer: tonic arousal during the 1-minute VR experience (which would favour the SF-DCM module in PsPM), or event-related arousal? Dominik
-
This software has moved to https://github.com/bachlab/PsPM. Please make sure to use the latest version available there, and please post all remaining questions on https://github.com/bachlab/PsPM/discussions.
-
PsPM has moved to GitHub; could you post your question there? https://github.com/bachlab/PsPM
-
Hi Maria thanks for reaching out. Could you kindly share your full script, either on Github (https://github.com/bachlab/PsPM) or via email? This sourceforge page is not serviced regularly any longer. Best wishes Dominik
-
Hi Anna it is not obvious that slow fluctuations of the pupil size would influence your GLM estimates at all. Korn & Bach 2016 and Korn et al. 2017 did not find that a high-pass filter improves the validity of the ensuing estimates. So you may also do without. (But perhaps there is a specific reason if filtering turns out to be better in your case.) Dominik
-
Hi Marleen sorry for the late reply. Separating SCR in short succession and determining their amplitude is a great application of PsPM. However, what you need (as with any peak-scoring analysis) is a clear idea of when the SCR occur. Often this is precisely mandated by the experimental timing, but in your case, the pre-stimulus anticipation and the dynamic nature of the stimuli appear to preclude such deterministic approach, and DCM which allows to estimate the timing from the data, could be more...
-
Dominik Bach posted a comment on discussion Help
Hi Tanja SCR amplitude or magnitude measures from peak-scoring analysis are sometimes square-root or log-transformed to render their distribution more "normal", similar to a link function in a generalised linear model. However, I am not aware of any systematic investigation under which conditions this is meaningful or necessary - and the choice often seems to be based on tradition. This is why PsPM does not transform the estimates and leaves this to the user. In peak-scoring (and DCM), the distribution...
-
Manual
-
.m file help texts
-
Manual changes
-
Strucutre GUI updates
-
Manual updates
-
Manual updates
-
pspm_trim Helptext
-
Hello Lauren Defining the timeunit as 'markers' only means that the timings given in the Epoch file or entered manually are marker numbers. Thus having timunit 'marker' does not mean that the timing is retrieved from the markerchannel. This functionality is not offered by PsPM. Your work around was correct. This means you need to define a Epoch file with the timing of the epochs (ie. marker numbers) or enter them manually. An Epoch file is a two-column text file with on/offsets or a .mat file with...
-
If there is no error, and the stimulus frequency is not below 0.5 per second, then I'd think the assumptions of the evoked model are not fulfilled because (a) either SCR have variable onset, as is common in fear conditioning or during long (> 1 s) stimulus presentations, and/or (b) the latency is different because of anticipation (resulting in shorter latency) or prolonged stimulus presentation (resulting in longer latency wrt stimulus onset).
-
Hi Ondrej consistently negative SCR are physiologically implausible and so I think it is always a good idea to check what is wrong. They are more often encountered in standard (not model-based) analysis, when a new SCR occurs during the decay of the preceding one; in PsPM they usually only occur when the timing is wrong. However, in your case it is true that any analysis method will have difficulties disentangling SCR to stimuli with less than 2 s SOA. There is a biophysical limit here - the system...
-
This is correct, Danny, hope it works out! Best, Dominik
-
Hi Daniel importantly the model should not be informed about the trial labels - and this includes the presence of a US. The absence/omission of a US should be treated as an event just as well, as it is suggested to elicit SCR. So you have one matrix for all CS, and one vector for all US/absence of US. Dominik
-
It's difficult to spot the error without seeing the batch script you are using (or the condition file). However, events in this approach do not correspond to conditions - but to events within one trial (across all conditions). DCM is not explicitly defined about conditions (even though you can specify them, but this info is just stored for later). E.g. event 1: CS (two-column vector), event 2: US/US omission (one-column vector).
-
Hi Danny thanks for your message. If your CS is as long as 7.5 seconds (is it?) then DCM would be your best bet for analysing the data, as the CS response is unlikely to occur time-locked to the CS (which is what GLM assumes). We use DCM already for 3.5 s CS duration. Specifying event duration in the GLM specification is inherited from fMRI - but it is not physiologically plausible. While CNS neurons can fire for extended periods of time, sudomotor nerves just don't seem to do that - they fire in...
-
This is correct although I'm not entirely sure about the AIC equation, but that is an issue outside PsPM. Dominik
-
suggestions in GUI help
-
grammar in GUI help
-
Hi Gianluca in the second script, you create a matlabbatch, but you don't actually run it. After creating the job, the last line in the loop should be pspm_jobman('run', 'matlabbatch'); You don't need the lines of code after AUC2 until AUC2_Job For the import, you don't need to define each file manually. In place of the data files specified with matlabbatch{1}.pspm{1}.prep{1}.import.datatype.labchartmat.datafile = {...} you can just specify a cell array of string with the filenames, using for example...
-
Hi Dan happy to look into this dependency error, but such bug fixes often take a bit. Meanwhile, no need to do the analysis manually, even without dependencies. Just define the desired file names in the batch script as strings, e.g. using sprintf and/or fullfile. If the input file is "xxx.asc', then the imported file will be 'pspm_xxx.mat', and the trimmed file 'tpspm_xxx.mat'. Best Dominik
-
Hi Dan hopefully the new year will not bring too much more headbanging :) I think the problem is that you bulk process several files for import and trim, but the SF module can handle only one file at a time. Apparently the dependency does not work here. I'll have to look into this. In the meantime, you can restrict yourself to processing just one file at a time, for all modules in the batch. I assume the dependencies will work then. Do let me know if this doesn't fix the problem. Best Dominik
-
H Jonas For a start, please have a look here: https://sourceforge.net/p/pspm/discussion/help/thread/addfd37e/#5d96 Do update me if you want me to expand on this, or have further questions. Best Dominik
-
You can run a GLM with one regressor per trial rather than per condition (ie. define 'conditions' that just contain one trial). In DCM you automatically get trial-by-trial estimates.
-
Hi Ambra PsPM offers no first-level statistical tests. The reason is that the data are strongly auto-correlated, and models to account for that (as e.g. in SPM for fMRI) can be quite complicated, or unreliable with just one data channel. As a surrogate, you can safely assume that trial-by-trial statistics are independent, and do statistical tests on this level after extracting them to e.g. in R or SPSS. Dominik
-
Small changes to Export function
-
Hi Dan all SCR analysis packages, including PsPM, high-pass filter the data before modelling or peak-scoring, thus removing negativ values. In addition, SF models subtract the minimum value per epoch before analysing the epoch. If the baseline period and post-zeroing data are in separate epochs, I see absolutely no problem comparing them - the estimated amplitudes of phasic SFs are not affected by this filtering/baseline correction. If the zeroing occured during an epoch, you would have to check...
-
Hi Dan (1) This sounds very reasonable. Some people also analyse SCL, but over such long intervals there is a strong peripheral component in this measure and evidence that this is influenced by central/psychological processes is at best patchy. (2) PsPM takes a hierarchical summary-statistic approach - it works on individual subjects and then allows to analyse parameter estimates from each subject. Create a batch script for one subject and modify it to loop over subjects. The Tutorial section in...
-
Status PsPM
-
Hi Lauren if you model two responses that are 4 s from each other, then as you say the first one will overlap with the second. In fact, the peak of the first response will occur during a time when the second one is initiated. Note, however, that this does not mean they are not at all separable. The second response will only show up in SCR 1.5-2 s after the second event, when the first one is already decaying. If you simulate data with two eSCR, you will see that if your subject's response perfectly...
-
Hi. If the design matrix looks ok in the review model menu, then there is little to worry about in terms of the onsets - this is exactly the same matrix used for model inversion. It's always a good idea to check predicted vs. observed and reconstructed responses. If there is something wrong, perhaps the timing did not match the experiment stimuli in the first place? In any case, I'm happy to look at it if you post it here or send via email. Dominik
-
Structure GUI feedback
-
Hi Claudia to troubleshoot the matlabbatch problem: can you save the variable matlabbatch in mat file, and try to open it in the matlabbatch GUI? Ideally this will show up that/where something is missing, or if it is a code bug. Best Dominik
-
adapt recon for SEBR
-
minor changes
-
Hi Claudia thanks for posting. I'll look into it. Perhaps you could send a sample pre-import file? Dominik
-
Hi. You can import these data by either (1) bringing both into one matlab or text file, or (b) importing them separately and merging in PsPM. The timestamp information is not required, PsPM will ask you for the sample rate instead. Your condition definition is correct if your time units are 'markers'. Hope this helps. Dominik
-
Hi Jonas in principle this kind of analysis is possible, and given the 8 s post-stimulus interval, also very meaningful. The power of the PsPM approach particularly for SCR is (1) to use the full data rather than just a peak, and (2) to account for previous responses that overlap with the current one. They will still overlap with a 13.5 s ITI. To fully exploit point (2), I would recommend importing the data in continuous time, but setting the non-recorded segments to NaN. In other words, you work...
-
Dear Jo it is quite possible that the amplitudes are actually much higher on some trials. If you want to combine SN burst dispersion and amplitude to work out the SCR amplitude, the formula is SCR(t) = a exp((t-m)^2/2s^2) x SCRF(t) where t is time, x the convolution operator, a is the SN amplitude, m the onset, s the dispersion, and SCRF the canonical skin conductance response function (found in pspm_bf_scrf_f). So you can either work out the maths or do it numerically. Best Dominik
-
Dear Jo 1) Yes. Note that if you have a flexible latency model, amplitude of the sudomotor burst does not directly relate to SCR amplitude (only together with burst duration/dispersion). 2) I'd check the dispersion here. Amplitude and dispersion are in some cases not clearly separable. If it is very small on these trials, it may be a good idea to set the minimum higher (e.g. 0.3) in pspm_init, or to constrain it to 0.3 (new option in the repository version of the software). 3) Not sure why you would...
-
PsPM status
-
backroom function
-
Jobs
-
Hi Ondrej thanks for the feedback. Regarding the handling of NaN stats by the contrast manager, this is a very good point, and thanks for brining it up. The feature that leads to these NaN values is quite new, and we haven't followed the consequences through the contrast manager. We'll take care of this shortly and provide an update. In the meantime, the easiest solution would be to simply do the contrasts outside PsPM in Matlab, by loading the model, extracting dcm.stats, and simply averaging within...
-
new tasks
-
Dear Ben great you could solve this. Let me reply nevertheless in the interest of other users: When using PsPM or SPM, make sure the toolbox you are not using is NOT on the path. In case of errors, manually remove all directories relating to PsPM or SPM from the Matlab path. Then add (only) the root directory of the toolbox you intend to use. Best Dominik
-
new tasks
-
new tasks
-
Hi Ben thanks again for highlighting this issue. A fix is now available on the code repository and will be included in the next release. Best Dominik
-
Hi Jo the problem with fit statistics on the participant level is that they don't tell you how good your inference is. For example, in Bach et al. 2013 you can see that the GLM fit (expressed as model evidence, i.e. accounting for increased model complexity) becomes better with more flexible response functions, but the inference on the underlying psychological process becomes worse. The same was observed in Staib et al. 2015. In other words, the more complex model is more plausible than the simpler...
-
Hi Mira I assume you specified the event timings in samples - but they should be in seconds. Thus, PsPM 'thinks' all trials are outside of your file. The brace indexing message you get is probably from misspecified return of the error within the matlabbatch, and you can ignore it. Best Dominik
-
small changes structure GUI
-
Hi Ondrej thanks for sending this. I tried to run your script, and I noted you had turned off the warning messages. After commenting out this line, I received a warning that the first data file could not be loaded. When I load this file into the workspace I note that the amount of data in channels 1-3 does not match the file duration specified in infos.duration. This makes the file invalid. All channels in PsPM need to have the same recording duration, and PsPM takes care of this during import. How...
-
Hi Isabel could you use the standard import for matlab files? The required format for the matlab file is simple and is is documented in the manual. This will convert your data into a standard PsPM file and the rest should follow. Hope this helps Dominik
-
Structure GUI
-
Hi David PsPM expects continuous data. This is particularly relevant for EDA data where the responses are rather extended in time (> 90 s) and usually overlap with the next trial even if the inter trial interval is very long. The power of PsPM is that it takes all these data into account - in fact, ISI data are very informative. Best practice would be to import one continuous file per participant. Using 4 marker channels will be a bit of a challenge in PsPM as it usually expects all information in...
-
Hi Ondrej would you mind sending one of the files that crashes, and send me the code with which you call DCM, for investigation? If you don't want to post data publicly, send via email to the address provided in the toolbox start up. Best Dominik
-
Hi Jo I think what happens in the DCM is that there is no response at the specified time, but the little wiggles in the data mean that changing the amplitude parameter a bit has a neglible impact on the fit. In this scenario, the estimated parameter can be equal to the prior (which is the amplitude of an average trial). Regarding your GLM, there should not be a problem to model (a) and (b) as they are sufficiently far apart and there is some jitter in the timing. The problem occurs if participants...
-
Hi Ben thanks for flagging this up. We're in the process of fixing it - I'll keep you posted here. Best Dominik
-
comments adapted in get_timing
-
Hi Melissa PsPM uses a response criterion only for counting the number of response in SF models (here, it is 0.1 mcS). For other applications, traditional SCR analysis often distinguishes between response amplitude (averaging after excluding non-responses) and response magnitude (averaging over all responses). The PsPM approach gives you what is called "magnitude" here. Sometimes, magnitude is computed after setting responses below the response criterion to zero. However, this does not make a big...
-
Hi Ambra yes, segment length is simply the time interval after the event onset that you want to extract, and yes, pupil size usually is back to 0 after around 10 s. Of course you can use shorter or longer segment length, depending on your data. Hope this helps Dominik
-
new option for DCM
-
Hi Jo you are looking into possibility (2): the response starts before, or within the 1 s after, the event. It is quite possible that subjects respond to the fixation cross - I've seen that before. You should be able to see this using extract segments. Can you increase the scale on the plots, or plot manually? Happy to assist if something does not work with this feature. You could also extract segments (or run the FIR model) wrt event (1) to disentagle whether participants are responding to the fixation...
-
Hi Ambra for GLM, the amplitude estimate is the reconstructed amplitude per condition. For DCM (to which the thread referrred originally), it is one of the several estimates found in the stats - they are recognisable by their names. Dominik
-
Hi Ambra the 'free' option for latency is supposed to work with startle eye blink data. The algorithm behind it does not work very well with SCR data when the ITIs are such that SCRs overlap (ie. ITI shorter than at least 30 s). Instead, varying SCR latency is better accounted for by using the time derivative of the SCRF, and reconstructing the response. This is the default option in PsPM, and evaluated in Bach et al. 2013 Psychophysiology. For brief stimuli, I would have thought that DCM is not...
-
bugfixes and helptexts
-
New option in pspm_get_timing
-
Hi Mira PsPM should represent imported Acqknowledge markers as time stamps. You can check this by importing and displaying the file (using the "display" function in the tools drop down menu in PsPM). You should see whether the markers are in the appropriate places. In case the import did not work, could you send a sample file (here or via email) for us to check? Best Dominik
-
Dear Ondrej I hope this will help you: (1) DCM runs on a trial-by-trial basis and is agnostic wrt to trial identity. If you specify trial names, they will merely be stored for later usage in the contrast manager. However, in the contrast manager, you can also manually select individual trials. In the contrast manager, you don't enter a contrast vector (unlike SPM) but select trials with the mouse. This can be tedious for DCM output. If you are familiar with matlab, I'd suggest using batch scripting...
-
Hi Giulia there are two points here: first, the distribution of residuals in a statistical test. Since DCM amplitude estimates (just like classical peak scoring measures) are bounded from below, they are not normally distributed. However, when you compute contrasts (average differences) between conditions, they will usually be approximately normal. The more serious point is whether you believe the estimates, and I share your view that it's better to check when some of the estimates are one order...
-
pspm_init documentation
-
Hi Suzanne Phasic heart-period responses are not (yet) well characterised, both within and outside of the model-based approach. I would suggest to use all basis functions that do not overlap with the next stimulus, but importantly, include all of them in the model even if you only plan to analyse a subset. plot the observed responses using the tool functions extract segments and segment mean, to get an idea how well the PsPM will fit the data GLM output of a model w/o derivatives represents, loosely...
-
Hi Suzy if you are interested in the evoked SCR to the tone, I think a GLM should give you the answer. The tones are short enough to count as "short", i.e. they are as long as most of the stimuli used to develop the SCRF (see Bach et al. 2010 Int J Psychophysiology https://www.ncbi.nlm.nih.gov/pubmed/20093150). To look at habituation, you can either model each event as a "condition" in the GLM, which will give you single-trial estimates. Or you model all events in one "condition" and add a parametric...
-
Hi Giulia single-trial estimates are easily obtained in GLM (if you find that GLM is suitable here). Just model 1 trial per condition, ie. you have as many conditions as you have trials.- I'm not sure GLM is the best way to analyse response to a 8 s stimulus. Are you modelling the stimulus onset? GLM assumes that responses directly follow the stimulus onset with constant latency which may not be the case. You can look at averaged responses per condition (using Segment Mean in the Tools menu) to check...
-
Hi Hannah negative numbers in the data are not an issue. PsPM treats the data to be on an interval scale, i.e. with no absolute zero. Offset (positive and negative) is fitted with an intercept in GLM, and for DCM the minimum in the data is subtracted before fitting. I guess it is ok to first median filter and then use the standard PsPM filters. The data still appear rather noisy, and I spotted some remaining artefacts. You can also mark them manually (using the data editor) and remove that from model...
-
Hi Ambra exactly - so this is slightly different from the way SPM works. If you have different conditions per session, it is a good idea to use Review Model to check the names of the regressors in the model. That way you can be sure you construct the contrast vector correctly. Dominik
-
Hi Ambra think of the contrast vector as a concatenated vector for all sessions - i.e. if session 1 has 36 trials, and session 2 has 11, then the 37th entry in the contrast vector refers to the first trial of session 2, and the 48th entry to the first trial in session 3. Hope this helps Dominik
-
Hi Roland GLM makes the assumption that the elicited SCR is time-locked to "something" that is known, e.g. an external event, or a subject's button press. It will miss (or mis-assign) SCRs that are not time-locked. DCM is a more flexible way of dealing with unknown timings of responses. If you expect more than 1-2 SCR per event (have a quick look at your raw data), the parameter space gets difficult to deal with, and then the SF model is more constrained (it is actually the same model but the dispersion...
-
Specify duration = 10
-
GUI -> tools -> extract segments
-
Hi Ambra the illuminance model is unlikely to be of much help here (unless you are interested in deconvolving the time course of neural input into the pupillary system, something that is not implemented in PsPM yet). So the question is, does the fear conditioning model fit appetitive learning experiments, too. Since we haven't explicitly tested this (and have no data to test it either), I'd suggest a quick fix. Plot the CS+ and CS- responses and their difference. If the difference curve looks 'similar'...
-
Hi Daria I understand. This is odd indeed. I agree that the temporal gap between CS and US could have an impact - it is possible that there is a response to the CS offset which is not properly modelled. If you want, I'd be happy to look at your new simulations and results, and it could be useful to look at the mean time course for the different conditions in your experiment. If you don't want to post this publicly, send me an email Best Dominik
-
Hi Daria my suspicion is that the simulated data are too different from what SCR data would look like. You write you used an HRF for simulations - is that a hemodynamic response function? This would have a different delay and spread than SCR. You can easily see this by plotting your HRF together with an SCRF, e.g. the one implemented in PsPM (pspm_bf_scrf_f). If you simulate data with this SCRF, then the model should fit (almost) perfectly. (Almost, because the non-linear model uses a slightly different...
-
Thanks for the message. We'll investigate this week. In the meantime (and actually more generally), I'd suggest exporting statistics into a more specific statistics software for the second-level analysis. This is more flexible, since the tests implemented PsPM are really just t-tests. Best wishes Dominik
-
Dear Ambra this is an important question, not only for PsPM. For most psychophysiologial measures, the dynamic range is limited, so when stimulations come too close in succession, the mapping from stimulation to physiological measure becomes non-linear, which is a problem for most analysis approaches, PsPM and otherwise (e.g. computing averages over trial-by-trial peak amplitudes). For SCR, we have specifically shown with intraneural stimulation that this mapping is linear up to about 1 stimulus...
-
Hi Ambra for the anti-alias filter, only a hardware filter makes sense. It will probably be IIR then. If there is no hardware low-pass filter, do not filter and oversample generously - 2k will suffice. No hardware high-pass filter is needed but make sure the signal does not exceed the converter's dynamic range with the gain you're using. Dominik
-
Dear Ambra I would recommend not using any filters during acquisition, only an anti-alias low-pass filter at half the sample rate. Any other filtering (including the 0.05 Hz high-pass filter) can be done off-line. PsPM expects non-filtered data by default. Dominik
-
Thnaks for letting us know - this is excellent news! Dominik
-
Yes, this should work. If you code your marker channel as binary with 1 and 0 this should work as well. Best, Dominik
-
Hi Mariya from the technical side this is easy - you can write your own RF as a matlab function and specify it as basis function for the GLM. The trickier part is how you develop and justify the RF. I guess you'd need an independent data set to do so. Hope this helps Dominik
-
PsPM marker channels contain time stamps of the marker events (with respect to file onset) in units of seconds. Dominik
-
In principle this should be possible - but they are represented as continuous data, so best to define markers at the epoch onsets to keep track of the epochs. However, if you want to model overlapping responses (depending on your modality), this could be slightly more involved and require some programming. Hope this helps Dominik
-
Hi Lauren DCM timing is always in seconds. GLM timing can be in seconds, samples, or markers. Regarding your notes on the manual: the units of a channe of type "marker" is given in the field header.sr - if it is 1 then the units are seconds. So the events in the DCM tutorial are read out from a marker channel that is in seconds, and the events are thus in seconds as they should be. Overlapping trials are not necessarily a problem, but depending on how much overlap and how many responses in the data,...
1 >