Cleveland and McGill (1984)

Figure 4, Cleveland & McGill (1984). Different types of comparisons with bar charts.

• Hierarchy of visual features:
  • Accuracy for length > area > volume
  • Experiments manipulated position and angle
    (no actual area/volume experiments)
• Type 1 vs. 3 comparisons: cognitive grouping

Zacks et al. (1998)

Common Graph Types. (Modified from Figure 2)

• Matching task: higher error on volume than area

• Estimation task: Error for volume > line > area

Illusions and Context

Developments in 3D Graphics

Developments in 3D Graphics

3D printed charts

Process

Integrated into the course to align with current topics

0. Informed Consent
1. Reflection: process of scientific research
2. experiment participation
3. Reflection: experiment purpose, hypothesis, error, variables, randomization
4. Reflection: what students learned from 2-page abstract
5. Reflection: How 15 minute conference presentation differed from abstract

Stimuli

Demographics

Demographics

Demographics

Data Cleaning

Analysis is of 5831 trials completed by 325 participants

Data Distribution

Figure 2: Number of trials across ratio and plot display type.

Post semester inventory indicates that some kits disappeared…

Results

Figure 3: Violin plots of response accuracy by true ratio value. The data is highly variable, but most responses are reasonably close to the true value. Rounding to 5 artifacts are clearly visible.

Results

Accuracy: \(\frac{\text{Estimated Percent} - \text{True Percent}}{\text{True Percent}}\)

Hypotheses

• “Students will get progressively less accurate as questions were asked”
• “Do students change their answers when asked the same question over and over?”
• “3D printed bar charts will lead to more accurate ratio judgments compared to 2D or 3D digital charts.”
• “That 2d is preferred over 3d. It cleans up the data presentation.”
• “The public can more accurately understand data when it is provided to them in a 2-D graph format.”

Sources of Error

• “Fatigue effect over the course of making many judgments, learning patterns from seeing the same ratios multiple times, possibly difference in eyesight among participants.”
• “There are no line values to help measure it when there is a small difference between the graphs.”
• “People that are guessing”
• “People in the sample misunderstanding directions.”

Elements of Experimental Design

• “I believe every graph was randomized for every student. I also believe the practice was meant to be a bit of the control group.”
• “Randomization was used as participants all received different sets of graphs to examine. A person wasn’t assigned to a certain set of graphs and it was up to random chance on which set of graphs they received.”

Elements of Experimental Design

• “I think that there is randomization but not a control group in my opinion, because there isn’t one group that is left alone or not studied….”
• “Randomization was not used because it was offered as an extra credit assignment in class.”

Conclusions

• No evidence of large differences between perception of 2D and 3D bars in this sample
  • Students may be less motivated to guess accurately
  • Students used to 3D effects may be less impacted by computer graphics
• Integrating graphics research into the classroom is fun for students! … but data may not be of the same quality as paid samples