feature request: diagnostic plots of entire timeseries + epochs

[shawntschwartz]

features/requirements:

• [x] change plotting mechanism to base graphics to remove ggplot2 dependency + speed up plotting!

• [x] allow step-wise interactive plotting of particular windows

• [x] paginate full time series plots to demonstrate effects of each step (overlay each step)

  for now, this functionality has been tabled to avoid needing to figure out flexible solutions for the limited interactive features provided by the R graphics window out of the box. perhaps this could later become part of an interactive shiny interface, but for now, it seems reasonable to take the approach of just plotting a subset of the timeseries for qualitative assessment (and then allowing the user themselves to take the raw/preprocessed timeseries and plot them however else they'd like on their own). -shawn (sept. 12, 2024)

• [x] allow these step-wise plots to be generated and saved out (especially if user is running analysis on a headless machine)

  similar to above, this can be a feature we can easily implement as a flag within the function, but doesn't seem like it's necessarily our responsibility to provide the additional scaffolding to save out these plots given; instead, we can relegate this responsibility to the end user, who can use their graphics function of choice (i.e., pdf(), png(), etc.) to save out these diagnostic plots as they so please to do. -shawn (sept. 12, 2024)

[shawntschwartz]

@mh105 can you provide any additional thoughts about the two points i raised in the initial issue comment above? thanks!

[mh105]

paginate full time series plots to demonstrate effects of each step (overlay each step)

I agree. Diagnostic plots and Plots of entire time series are two different concepts that serve different purposes. For a preprocessing package eyeris (at least for now) shouldn't have to worry about providing a nice pagination GUI to flip through the entire time series / all epochs. We may add this bonus functionality later, but for now it's sufficient to provide enough visualization for diagnostic evaluation.

My suggestion is to randomly pick 3 epochs (don't just show one because that doesn't feel representative enough) and show the step-by-step impacts of the processing pipeline. So you could have different steps across the rows, and 3 columns, or transpose it. This way the qualitative assessment is not just on a single epoch. And when people should be able to call the plot function again, which should randomly select yet another 3 epochs -> so if people want they can just do this however many times their hearts desire until they feel they have reached sufficient qualitative assessment. If the pipeline didn't epoch, just randomly pick 3 five-second periods and do the same thing. And you are right that users will have access to all the raw/preprocessed time series anyways, so they can plot them however they want. What do you think about this suggestion?

allow these step-wise plots to be generated and saved out (especially if user is running analysis on a headless machine)

Oh yeah no. Saving svg is easy enough in R that you don't really want to spend the efforts to accept and parse a saving directory and all that. People can and should be able to chain a saving call to the figure handle themselves, if they are comfortable enough to run the pipeline on headless machines.

[shawntschwartz]

great, thanks! glad we're on the same page about this @mh105!