Social Exp Sciviz Plots

[jkbhagatio]

To discuss the sciviz plots for the social experiments. Some thoughts here:

I was thinking we could have three types of webpages, for different timescales of the experiment:

1. For the overall experiment
2. For each light-cycle (every 12 hours)
3. For each block (every 1-3 hours depending on the sampled block time)

We could run a dash dashboard for each type of page, that contains plotly plots.

The key data features I think we'll want to plot for these Social experiments are:

• Subject Weight
• Subject Position / Ethogram
• Pellets consumed / feeding bouts
• Wheel distance spun
• Patch preference

I'm thinking of mainly having two types of plots on each page: "per-subject" plots, and "for-all-subject" plots

[jkbhagatio]

So if we go with this "multiple types of pages" approach, thinking something like this right now:

Block Page

• Header: "Arena: block_start_time - start_end_time"
• Plot 1: Patch stats: Patch rate for each patch as a bar, sampled pellet threshold values as overlaid scatter
• Plot 2: Weight: Weight over time
• Plot 3a: Position: X-Y line position over time with running cumulative total distance traveled
• Plot 3b: Position: X-Y heatmap
• Plot 3c: Position: Position ethogram
  • Contiguous blocks of time spent in given regions. Downsampled to 1s windows?
  • One row for each: nest, corridor, field, patch1, patch2, patch3; text overlay of total distance traveled within each continguous block?
  • Color-coded for when in the same location as 1+ others (one color per subject)
  • Text: percentage of time together overall, and by region
• Plot 4: Behavioral Ethogram
• Plot 5: Pellets: Pellets per patch - cumulative line plots, x-axis time
• Plot 6: Wheel distance: Wheel distance per patch - cumulative line plots, x-axis time
• Plot 7: Patch preference: Running, cumulative normalized patch preference: by wheel distance, and time spent
• Plot 8: "Learning" of block: Preference for High-reward and/or Med-reward > some_thresh

• Notes
  • By "plot" I mean "axis". How these different axes are arranged into figures is TBD.
  • Per subject plots: All except 1
  • All subject plots: 5::
  • For each patch plot, info should include patch number and patch rate
  • Each in the set of (each subject, each patch rate, each patch number) should have its own unique color
  • Wheel distance is the preferred of the three metrics (wheel distance, pellets, time spent) for patch preference

Light-cycle Page

• Header: "Arena: light-cycle_start_time - light-cycle_end_time"
• Plot 1: Block viz calendar: Horizontal bars with each block's duration, x-axis in datetime
  • Clicking on block takes you to that block page?
• Plot 2: Block viz histogram: Histogram of block durations
• Plot 3: Patch rate counts: split bar plot, 3 groups (patches) of 3 groups (rate counts)
• Plot 4: Weight (same as 'Plot 2' in 'Block Page')
• Plot 5: Position (same as 'Plot 3' in 'Block Page')
• Plot 6: Ethogram (same as 'Plot 4' in 'Block Page')
• Plot 7: Pellets per patch by patch number (similar to 'Plot 5' in 'Block Page')
• Plot 8: Pellets per patch by patch rate (similar to 'Plot 5' in 'Block Page')
• Plot 9: Wheel distance per patch by patch number (similar to 'Plot 6' in 'Block Page')
• Plot 10: Wheel distance per patch by patch rate (similar to 'Plot 6' in 'Block Page')
• Plot 11: Running cumulative normalized patch preference by patch number (similar to 'Plot 7' in 'Block Page')
• Plot 12: Running cumulative normalized patch preference by patch rate (similar to 'Plot 7' in 'Block Page')
• Plot 13: Running windowed normalized patch preference by patch number (similar to 'Plot 8' in 'Block Page')
• Plot 14: Running windowed normalized patch preference by patch rate (similar to 'Plot 8' in 'Block Page')

• Notes
  • For plots 7::, vertical bars at time of each block switch. Text in-between each vertical bar indicating patch rates for that block.

Overall Experiment Page

• Text info: Exp Name, epoch root paths, total recording time?
• Plot 1: Arena light-cycle calendar:
  • Clicking on item takes you to that Light-cycle page?
• Rest:
  • Same as Light-cycle Page?
  • Can index / x-axis by block-number instead of time in some cases?

[jkbhagatio]

Steps going forward for this approach:

For the overall experiment
For each light-cycle (every 12 hours)
For each block (every 1-3 hours depending on the sampled block time)

1. Build up "Block" and "LightCycle" Tables
   a. New Block table that contains Block_start to Block_end

   • From BlockState we can get these times

2. Write functions to generate some of the first easy plots, assume required args are block_start and block_end

[jkbhagatio]

Priorities for Block Page:

Data: find an ideal way to get per subject info (e.g. wheel distance spun per subject)

Plot: PatchStats: Patch rate for each patch as a bar, sampled pellet threshold values as overlaid scatter

Plot: Weight: Weight over time

Plot: Position:

• Two plots per subject: 1) X-Y over time; 2) Cumulative distance over time

Plot: Wheel distance: Wheel distance per patch per subject - cumulative line plots, x-axis time (try downsampling to 10 Hz)

Plot: Pellets: Pellets per patch - cumulative line plots, x-axis time

Plot: Preference: (priority on wheel data; could do either cumulative (running sum) or windowed (running avg) priority on cumulative)

• Plot 1: Based on wheel data: create synthetic null (similar to as in presocial), plot true preference for each patch compared to confidence bounds for synthetic null
• Plot 2: 3 axes (each pair of patches): visualize running preference for each patch vs the other

Plot: Ethogram: Contiguous blocks of time spent in given regions. Downsampled to 1s windows?

Patch_Stats, Weight, Position, Wheel_Distance, Pellets, Preference (wheel data) are plots to focus on before focusing on Overall Exp Page plots.