Theme 4 (Part Whole Relationships) Comments
Comments that apply to theme 4. Be sure to read the comments for “all themes” as well.
There was no equivalent to this theme last year.
I don’t even have good suggestions on what might be similar to suggest what to read. If you find something, please put it into the Canvas discussion!
Some of the thoughts came up from specific groups, but apply to everyone (see #10,14). Generally, in this category most of the comments apply broadly.
Many proposals had reasonable sources of inspirations. Please post things on the Canvas discussion!
Specific number comments given to at least one student (these would appear as 4.X in your list). Again, the list given in the feedback may not be exhaustive (i.e., some of these might apply to you even if they weren’t listed in what was sent to you).
Good plan to prototype on paper first and do extensive explorations, and only to implement afterwards.
When evaluating/critiquing/picking, you will need to consider tasks (what are the different designs good/less good at). Be sure to include some of the baselines (e.g. a bar chart with error bars per category).
Focusing on the static case is fine - especially if you do a good job at coming up with interesting/effective designs. Considering how you might extend a design that works statically (for example with hovers) can be an add-on.
Please do document your exploration. A catalog of ideas that should not be used can still be helpful - it’s good to know what not to do!
Focusing on comparing groups is good - but remember, the groups still have diverse distributions, you don’t just want to give the mean for each group. (although that might be a start)
Hopefully you will get to try out your designs on real data.
Using what you create to tell a story is good - but hopefully that is a result of having good ways to show distributions.
More than average is a good start. It isn’t clear what higher order statistics are useful, and how the constraints (of summing to a whole) affect them.
I like representative sampling as a potential idea. Complexity in how to choose the samples as well as how to show them.
Tasks might be specific or general. If you have a task specific to ATUS, there is probably a generalization of it broadly. If a task is general, there is probably a specific interesting example using the ATUS data set.
I like the idea of a space of how well a design shows part/whole and distributions. To my knowledge, existing designs do one or the other (so they would be along the axes). But hopefully you can generate examples that are more in the heart of the graph.
Don’t feel compelled to implement all designs, but to document them (see 4). If you can determine something is good/bad from a sketch, you might not need to waste time implementing.
Focusing on tasks is great! Please make sure to document this.
Working on smaller subsets is sufficient. You don’t need to only summarize the whole set - even small groups can have interesting distributions that are challenging to show. (the group that got this comment explicitly talks about this)
Good motivating examples are useful. If you have one, (e.g., you got this comment), please post it to the Canvas discussion so it can inspire/challenge others.
I didn’t think through specific designs given, but they are (at least) good starting points. Keep points 2 and 4 in mind.
Remember, the goal is to show the distributions of the part whole relationships. The goal should be methods for showing this.
If you do try to tell stories using what you develop, try to focus on stories where the distributions turn out to be important to the story. (e.g., simply using the average or median isn’t good enough to tell the story)
Starting with a small number of categories and then generalizing is a reasonable strategy.
Doing an informal “study” (asking your friends/classmates) is a good idea - but might be ambitous to do in a thorough way.