|
James Scott-Brown
@
jamesscottbrown
|
|
Software Engineer at @Oxford_NDPH. Previously D.Phil student in Synthetic Biology @UniofOxford and @synbioCDT, then RA in Visualization at @ImperialDSI.
|
|
|
549
Tweetovi
|
402
Pratim
|
201
Osobe koje vas prate
|
| Tweetovi |
|
James Scott-Brown
@jamesscottbrown
|
23. sij |
|
And yet there is no recommendation against "skin sticks".
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
23. sij |
|
One solution is to move the y-axis to the right-hand side, label the x-axis "Finding this much of the chemical network requires...", and label the y-axis "Log N experiments".
I think readability would be improved by plotting N on a log-scale, rather than log(N) on a linear scale
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
22. sij |
|
I think I saw a nice study of this, but can't find the reference right now.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
22. sij |
|
Judging the vertical thickness is not quite as easy a task as judging the length of parallel lines/bars. People don't necessarily judge the thickness in a vertical direction: they can be misled by looking at the thickness as would be measured horizontally or at an angle.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
22. sij |
|
It's a ("length proportional") "linear diagram" with coloured overlaps: napier.ac.uk/~/media/worktr…
I don't know of many tools for drawing them, but there is cs.kent.ac.uk/people/staff/p…
I generated this essentially from scratch using D3.js; I can share the code if that would be helpful
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
19. sij |
|
From Gaither's Dictionary of Scientific Quotations: pic.twitter.com/balRhVG6IU
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
18. sij |
|
Randomising the order of the items hides a lot of structure that could be seen if items belonging to the same combination of sets were positioned contiguously: pic.twitter.com/qvQxrnngjI
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
17. sij |
|
The sign up link seems to have been truncated: I think it should be forms.gle/afjxV4Hv6qsdww…
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
16. sij |
|
Direct link to this page: hdl.handle.net/2027/uc1.c0000…
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
10. sij |
|
If a pitched article is accepted, when would the finished article be due?
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
If a system provides a sufficiently elegant and concise way to specify graphics, you can use these specifications without needing new names as a shorthand, and users can directly specify encodings rather than being limited to picking from a chart catalogue.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
(Also, I noticed that Pictura cites Wickham as the 'first grammar of graphics' - what about Leland Wilkinson's The Grammar of Graphics?)
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
Such an approach avoids the need for many names: there is no notion of a barchart that needs naming, just a mapping from data attributes to the visual attributes of bars (a nominal data attribute -> horizontal position, and a quantitative variable -> length).
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
*Naming* specific types of charts seems orthogonal to creating a grammar in which they can be expressed.
Whilst Principia Pictura includes a view composition algebra, it seems to be built on a taxonomy of base charts, rather than using a grammar to express the visual encoding.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
I understand the importance of formalization, but am less convinced about the importance of names.
A system can generate draw a bar chart of average GDP, and then generate an appropriately descriptive subtitle without needing a new name that's more specific than bar chart.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
When plotting a barchart, what parameter would you assign a value that depended on whether the y-axis represents counts/frequency or a summary (e.g. mean)?
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
Calling something a 'summary plot' rather than a 'bar chart' isn't helpful, because just telling me the plot/chart type doesn't tell me what variables are being plotted - and if you tell me this, then I don't need a distinction to be made by the chart name.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
4. sij |
|
It's a bar chart. Specifically, it's a bar chart *of average GDP per capita by continent*.
Bar charts of summary statistics don't have or need a dedicated name, because the fact that they show summary statistics is communicated by the description of what is plotted.
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
31. pro |
|
This is very nice. As a refinement, have you considered using a small diagram in place of the color legend - e.g., something like this? pic.twitter.com/TfSX250mom
|
||
|
|
||
|
James Scott-Brown
@jamesscottbrown
|
29. pro |
|
I agree with @NElmqvist's arguments:
twitter.com/NElmqvist/stat…
|
||
|
|
||