A Comparison of Visualizations for Identifying Correlation over Space and Time

Supplementary Material

Vanessa Peña-Araya, Emmanuel Pietriga, Anastasia Bezerianos

Inria, Univ.Paris-Sud & CNRS, Université Paris-Saclay

InfoVis 2019

Sample trial images

Study Material

The following items link to the files we used to conduct our experiment:

• Folder datasets_generation:
  • data_generation_process.pdf: gives an overview of the process of dataset generation with the functionalities of each Python class.
  • data_states_config.py: stores the spatio-temporal configuration for a pattern of the 8 tasks considered in the study, stored in the class StatesDataConfig.
  • data_generator.py: given an instance of StatesDataConfig, generates the final dataset distribution of correlated data.
  • data_exporter.py: helper classes for exporting to JSON or CSV.
  • data_generator.ipynb: Jupyter notebook with the generation of the 80 datasets for each of the 24 main trials.
  • obesity.json: helper file to export datasets in JSON.
• Folder datasets contains the final 80 datasets in CSV and JSON.
• Folder images_generation:
  • HTML files: contain scripts in D3 or Vega for the generation of visualizations used in the experiment. They output individual images to be concatenated later.
  • Folders static and data: contain the necesarry files to run scripts in HTML files.
  • barchart.R and dorling_symbol.R: the R scripts used to concatenate individual images.
  All images have a width of 2700 pixels and are later resized to a lower resolution (1350 pixels in width) in order to minimize the delay when loading trials in the Web interface.
• Folder images contains the final set of 80 datasets x 3 visualizations = 240 images.
• Folder training slides (PDF) used during the participants' training phase.
• Folder experiment:
  • questions_training.csv and questions.csv: contain the questions for the 8 training trials and for the 24 main trials.
  • answers.csv: possible answers for each type of task.
  • user_configurations.csv: detailed configuration of the order of visualization and dataset config (column question_config).
  • user_question_configurations.csv: shows complete configuration of each trial assigned to each user.
  • user_answers_ms.csv: answers, self-perceived confidence and time taked for each trial for each user.
  • user_postquestionaire.csv: strategies and self-perceived easiness to complete the tasks for each visualization for each participant.
  • Background Information.csv: participants' demographics.
  • Final post questionaire.csv: post-hoc questionaire in which each visualization was ranked after all 3 sessions.
• Folder analysis:
  • Detailed_CIs.pdf detailed CI values used for our analysis (Completion Time and Error Rate).
  • user_answers_ms.csv -> experiment results used for analysis (same as in experiment).
  • A set of R scripts to be ran in order for each measure:
    • 1_aggregate_data.R aggregates (common for all measures)
    • files of the type 2-*.R that report means and CIs for all tasks (2-tasks_all_CI_TIME.R, 2-tasks_all_CI_ERROR.R, 2_tasks_all_CONF.R)
    • files of the type 3-*.R that report means and CIs for per-task analysis (3_tasks_separated_CI_TIME.R, 3_tasks_separated_CI_ERROR.R, 3_tasks_separated_CONF.R)
    • CI-Functions-Bonferroni.R includes the analysis functions. Bonferroni corrections were not applied in our analysis.
  • Analysis-README.pdf file with further information about task naming.