Configure a linear or logistic regression in Insights

Insights allows you to configure regressions on your data. Regressions can only be added to scatterplot charts - read this article for more information on configuring your Insights chart type.

Configure a linear regression

1. Select the “Scatter plot” chart type, and specify the appropriate x- and y-axes
2. Under the “Analysis” tab of the chart configuration, choose “Linear” as the regression line type
• Note: The linear regression is calculated using the y=mx+b model (m = slope; b = y-intercept) via ordinary least squares
3. Optionally choose an “Aggregation” option to display an aggregated view of y-axis values that share the same x-axis value
• If an aggregation is chosen, an “Error bars” option can also be selected to display either the standard deviation or standard error
4. On the top right of the chart, you can view the parameters and summary statistics of the calculated regression curve.
• Click on the information icon to view these values in a table
• Click on the “Copy values” icon to copy these values

Configure a logistic regression

1. Select the “Scatter plot” chart type, and specify the appropriate x- and y-axes
2. Under the “Analysis” tab of the chart configuration, choose “4-parameter logistic” as the regression line type
• Note: The regression is calculated using 4-parameter logistic model and the least squares fitting algorithm. The model can be found by clicking on the “Settings” icon next to the regression dropdown.
3. Click on the “Settings” icon next to the regression dropdown to view information about the model and fitting algorithm used
• Navigate to the “Parameters” tab of the settings modal to constrain the parameter values of the model. You can fix the parameters either to a specific value or to a range (greater than, less than, between, or equal).
• Under “Advanced settings,” you can optionally set the initial parameter values used for the model. For nonlinear regression, each parameter in the model must have initial values in order to improve the fit of the model to the data. Benchling can automatically set these initial parameter values for you. Poor initial parameter values may lead to a suboptimal fitting.
4. Optionally choose an “Aggregation” option to display an aggregated view of y-axis values that share the same x-axis value
• If an aggregation is chosen, an “Error bars” option can also be selected to display either the standard deviation or standard error
5. See this article for more information on outlier detection, which is only available on logistic regressions
6. On the top right of the chart, you can view the parameters and summary statistics of the calculated regression curve.
• Click on the information icon to view these values in a table
• Click on the “Copy values” icon to copy these values
• Note: For non-linear models, R² does not necessarily capture goodness-of-fit. We also include a p-value for how likely it is that the observations occurred by random chance, rather than following a logistic distribution. Lower p-values imply a more statistically significant model fit.