I want to record the time to execute or imagine a finger-tapping task, but the duration varies per trial. I watched AFNI bootcamp and think that maybe I should use the dmBLOCK response model (If different stimuli in the same class 'k' have different durations, but in my case, the same stimuli have different durations which's difference may be more than ±TRs), so is it appropriate to use dmBLOCK?
If use dmBLOCK, the duration is married to each stimulus time, all of this correlates to blocks not to trials. Is that right? My experiment is a two within-individual factors design. Total of four conditions (The duration of the task varies for different conditions, exp. the duration of condition one maybe 2±1s, ..., condition four maybe 30±5s，). I set up 30 trials for each condition, so how should I proceed with my experiment?
assumption one is 30 trials for condition one, 30 trials for condition two, then three and four, each condition was balanced between subjects using the Latin square. So there is only one block per condition. The problem is that the block duration of condition one is very short and the condition four is too long even more than 1,000 seconds. There must be something wrong here.
assumption two is each block contains less trials. Each block has a fixed number of trials, maybe 10 trials of each? Or 15 trials per block of condition one because of shorter duration, 5 trials per block of condition four because of longer duration. This way the number of blocks per condition is different, how to control each block is randomly presented while blocks are scaled differently. If the latter is better, how to properly set the duration of block? Is there some empirical criteria for this, I only knew that AFNI bootcamp mentioned that it's not to set less than 10s.
Finally, In the block design, do the intervals between trials need to be randomized? And how long is the rest set between block and block?
I am very appreciate that if there have any answers！
To accommodate varying trial durations, consider using dmUBLOCK in place of dmBLOCK, as detailed here.
If you are still in the experimental design phase, it would be beneficial to clarify the two scenarios you mentioned. For instance, are trials nested within each block? It is crucial to formulate your effects of interest before designing the experiment. For example, with a duration of 2±1s for condition one and 30±5s for condition four, are you comparing these conditions? If so, do you plan to compare condition one at its average duration (e.g., 2s) and condition four at its average duration (e.g., 30s)?
I thought about Q1 and Q2, and as you said, it's better to use dmUBLOCK model. As to the question in your reply, I may still be puzzled for my experimental design.
There are two factors in my experiment including the type of task and the difficulty of the task, they each have two levels respectively (MI: motor imagery, ME: motor execution) x (easy task, difficult task). As mentioned above, I want to record the time of task, but the duration varies per trial within each conditions. And duration varies widely between conditions (e.g., MI_easy about 2±1s, MI_diff about 25±5s, ME_easy about 3±1s, ME_diff about 30±5s).
The effects I am interested in are the main effects of task type and task difficulty and the interaction effect between them (including simple effects).
Since I first dealt with the duration modulation, I'm a little overwhelmed with experimental design. I'm thinking about block design. "Are trials nested within each block?", I think so.
My doubt is how to present the experimental stimuli?
One block for each experimental condition, and trials nested within block? Then the order of presentation should be MI_easy block--MI_diff block--ME_easy block--ME_easy block with each conditional block containing 30 trials. If so, does the time.1D file (e.g., 10:1 40:2 70:3) refers to the trial onset and duration? Does the intervals between trials need to be fixed or jittered?
Or, there are multiple blocks for each condition, the total number of trials for these blocks is 30 (e.g., one MI_easy block contains 5 trials, and the MI_easy condition would have 6 blocks). Then the order of presentation should be MI_easy block--MI_diff block--ME_easy block--ME_easy block--...--MI_easy block. If so, does the time.1D file refers to the block onset and duration?
If, according to the first scenario, the response model corresponds to a single trial (maybe like event-reated? but all trials are the same within block). If second, the response model corresponds to blcok (which has multipe closely-spaced activations). That's what's so confusing to me. I don't know much about this, so I'm asking very sincerely if you can give some advice.
I have two follow-up questions regarding the experimental design:
Since the four experimental conditions have different intrinsic durations, the resulting hemodynamic responses will also differ in magnitude and duration. Are you primarily interested in comparing the magnitude of the responses, or both the magnitude and duration?
What is the reason for grouping trials of each condition into blocks? Is this due to an experimental constraint? In other words, what prevents you from designing an event-related experiment with fully randomized trials? Or, is the goal to compare the conditions at the block level rather than the trial level?
Hi again. In fact, both the magnitude of the responses and duration are important.
I first thought of the block design because of its signal-to-noise ratio, no cues needed for different condition of stimulus, no cognitive impact between task switch and so on. I have not rejected the design of an event-related experiment with fully randomized trials.I'm confused about my experimental conditions in both block and event-related design.
As above, one factor is task difficulty. Easy level is about 2 or 3±1s, diffcult level is about 25 or 30±5s.
If use event-related design, the presentation time of diffcult level trials is too long (for my limited understanding, in an event related design, each task is presented individually for a short amount of time –e.g. 3s.). Is it okay to present about 30 seconds for a trial?
If use block design, each single trial at the difficulty level can be viewed as a block (because of the continuous tapping for about 30 seconds, e.g., like the task phase in Figure C). But, I knew that the blocks should be greater than or equal to 10 seconds (>=10 s). Does this mean that each single trial at the easy level can't be viewed as a block. What I understand is so limited.
Sincerely,
Lin
(An experimental design procedure diagram for a motor imagery study)
I apologize for my limitations in providing more specific suggestions due to my unfamiliarity with your research domain. However, I believe it might be challenging to simultaneously focus on both the magnitude and duration of the response in a condition-level comparison. For instance, an easy task might elicit a similar response magnitude as a difficult task, but the duration difference could simply be due to the inherent length of the latter task.
If you decide to focus solely on comparing response magnitude, it might be reasonable to consider adopting a block design where all blocks have approximately the same duration. In other words, easy blocks would have more trials (e.g., 10-15) than difficult blocks (e.g., 1). Regardless of your approach, conducting a pilot study with a small group of participants might be highly beneficial. This would allow you to test and compare the validity of one or more experimental designs before committing to a reliable one.
Hi. Thank you for your patience in replying! I have a few final questions.
As you say, the duration difference could simply be due to the inherent length of the latter task. It's more reasonable to focus on amplitude. When all blocks have approximately the same duration, each conditional block contains a different number of trials.
If each condition has the same number of blocks, the easy condition will have more trials than the diffcult one. Should I set the number of trials to be the same for each condition, or the number of blocks? If trials, preferably no less than 30 trials for each condition, right? If blocks, no less than how many is better? And Is it necessary to set jitter for each trial in the block in this experiment?
If your focus is specifically on comparing the hemodynamic response magnitude across conditions rather than their duration, I recommend setting those conditions with an equal number of blocks. To ensure duration equivalence across conditions, consider incorporating more trials (e.g., 10-15) in easy blocks compared to difficult blocks (e.g., 1).
It's unnecessary to introduce jitter between consecutive trials, and you can include a minimal or no inter-trial interval based on the nature of the experiment. Again, conduct a pilot study to test out the design before fully committing to it.
Thank you so much for replying, it helped me so much, thanks again!
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.
