Abstract:Parallel coordinates plots (PCPs) are a widely used visualization method, particularly for exploratory analysis. Previous studies show that PCPs perform much more poorly for estimating positive correlation than for estimating negative correlation, but it is not clear if this is affected by the aspect ratio (AR) of the axes pairs. In this paper, we present the results from an evaluation of the effect of the aspect ratio of axes in static (non-interactive) PCPs for two tasks: a) linear correlation estimation and b) value tracing. For both tasks we find strong evidence that AR influences accuracy, including ARs greater than 1:1 being much more performant for estimation of positive correlations. We provide a set of recommendations for visualization designers using PCPs for correlation or value-tracing tasks, based on the data characteristics and expected use cases.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the impact of different aspect ratios (AR) of axes on the task accuracy when parallel coordinates plots (PCPs) perform specific tasks. Specifically, the author focuses on how AR affects the accuracy of the following two tasks:
1. **Linear Correlation Estimation**: Evaluate the accuracy of users in estimating the linear correlation between variables under different ARs.
2. **Value Tracing**: Evaluate the accuracy of users in tracing specific values among multiple axes under different ARs.
### Research Background
Previous studies have shown that PCPs perform far worse in estimating positive correlations than in estimating negative correlations, but it is not yet clear whether this difference is affected by the aspect ratio (AR) of the axes. Therefore, this paper experimentally evaluates the impact of AR on these two tasks in static (non - interactive) PCPs.
### Main Contributions
1. **Evaluating the Impact of AR on Task Success**: The study found that AR does affect the success rate of tasks, and the magnitude of the impact depends on the sample size and AR.
2. **Asymmetric Impact**: The impact of AR on correlation estimation is asymmetric, that is, the estimation of positive correlations is more accurate when AR is greater than 1.
3. **Design Recommendations**: Based on data characteristics and expected usage scenarios, specific recommendations on how to set AR are provided for visualization designers.
### Research Methods
- **Task A: Correlation Estimation**: Participants were required to identify the correlation between a pair of adjacent PCP axes and judge it as strong positive/negative correlation, moderate positive/negative correlation, weak positive/negative correlation, or no correlation.
- **Task B: Value Tracing**: Participants were required to estimate the position of a specific value among multiple axes on the last axis.
### Experimental Design
- **Sample Selection**: Datasets with different sample sizes (40 and 160) were generated, and each dataset contains 4 variables.
- **AR Selection**: Five AR values (0.25:1, 0.5:1, 1:1, 2:1, 4:1) were selected, and the effectiveness of these AR values was verified through pre - experiments.
- **Task Structure**: The task was divided into two parts to reduce the influence of the learning effect.
### Results
- **Task A**: The results show that AR significantly affects the accuracy of correlation estimation, especially for positive correlations, which perform better when AR is greater than 1.
- **Task B**: The results show that AR significantly affects the accuracy of value tracing, with the best performance when AR is 1, and the worst performance when AR is 0.25 and 4.
### Discussion
- **Correlation Estimation**: AR has a large impact on correlation estimation, especially when the sample size is small. AR greater than 1 helps to estimate positive correlations more accurately.
- **Value Tracing**: The impact of AR on value tracing is also significant. The best performance is achieved when AR is between 0.5 and 1, and an AR that is too small or too large will reduce the accuracy.
### Recommendations
- **Correlation Estimation**: For low sample sizes, it is recommended to use AR greater than 1 to balance positive and negative correlation estimations; for unknown sample sizes, AR = 2 may be the most balanced choice.
- **Value Tracing**: It is recommended to use AR between 0.5 and 1 to improve the accuracy of value tracing; avoid using very small (<0.25) or very large (4) ARs, unless combined with other interactive methods such as brushing or axis - ordering techniques.
Through these studies, the author provides important guidance for the design and use of PCPs, especially in exploratory analysis.