Modeling Just Noticeable Differences in Charts

Supplemental materials for the paper on Modeling Just Noticeable Differences in Charts accepted in IEEE VIS'21.

paper link: https://deardeer.github.io/pub/VIS21_LSD.pdf

'stimuli_data' folder

the bounds for conditions, meta data driven 7200 plots

- *_bounds.csv // (epsilon_lower epsilon_upper) for each conditions in three charts
- plot_data/*_test_*.json // data for the 7200 plots, i.e., heights of bars (in pixel) for bar chart, radius for bubbles in bubble chart, angles of fan in pie chart
- plot_data/*_meta_*.json // meta data for the 7200 plots, i.e., the 'intensity of standard stimulus' (baseHeight/baseR/baseAngle) in each plot, 'difference' (delta*), and 'distance' (gap*)

"jnd" folder

the estimated jnds, sorted by chart type

- jnd_bar.csv
- jnd_bubble.csv
- jnd_pie.csv

Data dict:

• userid: [String] unique user id
• intensity: [Int] intensity of standard stimulus (i.e., height in bar chart, radius in bubble chart, degree in pie chart)
• distance: [float] distance between standard stimulus and compared stimulus
• jnd: [float] estimated Just Noticeable Difference
• pse: [float] estimated Point of Subjective Equality

"rawdata" folder

the raw data collected for each trial, for each chart type, the raw data sorted by user, in file name formate of "userid.csv"

- bar/*.csv  
- bubble/*.csv
- pie/*.csv

Data dict:

• yes-answer: [Boolean] 1 if answering the compared stimulus is larger/taller than standard stimulus; 0 if not
• time: [Float] time cost of answering this trial- intensity: [Int] intensity standard stimulus in this trial
• distance: [float] distance in this trial- difference: [float] the intensity difference between compared stimulus and standard stimulus, if negative, it means compared stimulus is smaller/shorter than standard stimulus.

"stat" folder

the yes-answer counting for each condition, for each chart type, the yes-answer is sorted by each user, each condition, in file name formate of "userid_distance_intensity.csv"

- bar/*.csv
- bubble/*.csv
- pie/*.csv

"stat_dest” folder

computed JND for each user at each condition, in each chart type, in file name formate of "userid_distance_intensity.csv.txt”

- dest_bar/*.txt
- dest_bubble/*.txt
- dest_pie/*.txt

Date dict:

• X: [Float] the intensity of compared stimulus
• Yes: [Int] the number of yes-answer in that 10 trials
• Total: [Int] the number of trials in that comparison, we repeat 10 trials, so it is 10

“code” folder

the python / R script that we run to analysis the data

-draw_jnd_distribution.ipynb // draw the box plot distribution in Fig. 4
-compJND_2021_submitted.R // read data from stat and compute JNDs
-jnd_*2021_submitted.R // JND modeling in three charts

To cite this work

@article{lu2021jnd,
  title={Modeling Just Noticeable Differences in Charts},
  author={Lu, Min and Lanir, Joel and Wang, Chufeng and Yao, Yucong and Zhang, Wen and Duessen, Oliver and Huang, Hui},
  journal={IEEE Transactions on Visualization and Computer Graphics},
  year={2021 (to appear)},
  publisher={IEEE}
}