Transforming Datasets to Stack Charts

A couple of quick ideas about visualizing data, especially with regard to comparison.
Author
Affiliation

Harold

TechAscent

Published

June 24, 2025

Keywords

charts, datasets

With observed data, presumably from two runs of some experiment…

(def ds0
  (ds/->dataset "https://gist.githubusercontent.com/harold/18ba174c6c34e7d1c5d8d0954b48327c/raw"
                {:file-type :csv}))
(def ds1
  (ds/->dataset "https://gist.githubusercontent.com/harold/008bbcd477bf51b47548d680107a6195/raw"
                {:file-type :csv}))

Well, what have we got?

ds0

https://gist.githubusercontent.com/harold/18ba174c6c34e7d1c5d8d0954b48327c/raw [500 1]:

y
-0.09138541
0.73573478
0.66637442
1.42894310
1.17985915
2.10245096
2.35628501
1.65951387
2.66932952
1.96287689
…
6.30911743
6.65394635
5.88407917
6.59312352
6.32078823
5.78220740
6.11383638
6.62701870
6.29688536
5.87255145
6.34171349

A few hundred numbers… Hm…

ds1

https://gist.githubusercontent.com/harold/008bbcd477bf51b47548d680107a6195/raw [500 1]:

y
1.23590349
0.97176804
1.44779983
2.09836076
2.39260885
2.33861635
2.55252144
2.75108032
3.42274612
3.13478376
…
22.45761328
22.35632666
21.93285307
22.24006990
22.51064120
22.38858256
22.53949283
22.57957379
22.31971585
22.69953383
22.23848485

This neglects the hundreds of thousands of years invested in evolving a visual system…

(-> ds0
    (plotly/base {:=title "Run 0"})
    (plotly/layer-point {:=y "y"}))
(-> ds1
    (plotly/base {:=title "Run 1"})
    (plotly/layer-point {:=y "y"}))

Better; however, our aim is to compare them… Which is higher?

(-> (ds/concat (assoc ds0 :v "Run 0")
               (assoc ds1 :v "Run 1"))
    (plotly/base {:=title "Comparison Between Runs"})
    (plotly/layer-point {:=y "y"
                         :=color :v}))

Now it’s up to the viewer to decide whether they like higher numbers or not.

There are a couple of interesting ideas in that last bit of code:

  1. assoc a constant onto a ds creates a constant column
  2. :=color takes care of grouping the results and the downstream display

Neat.

source: src/scicloj/tableplot/ideas/stacking.clj