Before exploring your imported data on the platform, reviewing potential correlations between different entities can help accelerate your findings. To begin, navigate to the “View Correlations” tab in the “Visualize” section.
How to Find Correlations
From the View Correlations page, you’ll find several options for analyzing relationships between data points:
- Inputs/Outputs: View correlations between selected inputs and outputs.
- Calculations/Outputs: Analyze correlations between calculated values (i.e. Solids %, density) and outputs.
- Outputs/Outputs: Discover correlations between different output data points.
- Component Outputs/Outputs: Compare component outputs for correlations.
Selecting Inputs and Outputs
To view correlations, select an input from the dropdown, or type the first few letters to speed up your search. Then, choose the relevant outputs you want to compare. You can also click “Show All Outputs” or “Show Outputs in Active Spec” to automatically include all relevant outputs.
Once you’ve selected the inputs and outputs, the platform will display correlations that meet the following criteria:
- Correlation strength above 0.5
- P-value below 0.005
- Consideration of minimum sample size
The result is a Sankey diagram, with distinct correlation lines for each output. A Sankey diagram is a flow chart where the width of the connecting lines (also called “flows”) represents the strength of the relationship between two elements—in this case, an input and its corresponding outputs. When reading a diagram, note the following:
- Nodes: The input (Polymer A) is on the left, and outputs (pH, Specific Gravity, etc.) are on the right.
- Flows: Bands connecting the input to outputs represent correlation strength:
- Thicker bands = Stronger correlation.
- Thinner bands = Weaker correlation.
- Colors: For example, pH: 60.00 °C shows a strong positive correlation with Polymer A (blue), while S @ 60s: -30.00 °C shows a negative one (red).
- Blue = Positive correlation.
- Red = Negative correlation.
By clicking any correlation line (flow), you’ll be redirected to the Explore Data tab to continue your analysis. In the example below, clicking on the pH: 60.00 °C output redirects to the Explore Data page, where the X and Y axis are pre-populated. For more information, refer to Explore Data.
You can also switch to the Correlation Matrix tab for a quantitative view of the results.
Additional Features and Settings
Scrolling down, you’ll find a list of all correlations for your selected dimensions. Clicking “View” will show the corresponding Explore Data plot.
Several settings are available to help refine your results:
- Minimum Samples: Adjust this to require a higher number of data samples for a more reliable correlation.
- Correlation Threshold: Increase this to filter out weaker correlations and focus on stronger relationships.
You can also use the “Show Outputs Without Conditions” option (via the cogwheel) to see outputs without specific condition parameters, giving you a broader overview of correlations.
Advanced Options
Finally, under “Advanced Options” (in the cogwheel menu), you can customize the correlation type (e.g., linear, quadratic, logarithmic) and choose whether to include outliers in your analysis.
