Setting Options for the Balance Node

Record balancing directives. Lists the current balancing directives. Each directive includes both a factor and a condition that tells the software to "increase the proportion of records by a factor specified where the condition is true." A factor lower than 1.0 means that the proportion of indicated records will be decreased. For example, if you want to decrease the number of records where drug Y is the treatment drug, you might create a balancing directive with a factor of 0.7 and a condition Drug = "drugY". This directive means that the number of records where drug Y is the treatment drug will be reduced to 70% for all downstream operations.

Note: Balance factors for reduction may be specified to four decimal places. Factors set below 0.0001 will result in an error, since the results do not compute correctly.

• Create conditions by clicking the button to the right of the text field. This inserts an empty row for entering new conditions. To create a CLEM expression for the condition, click the Expression Builder button.
• Delete directives using the red delete button.
• Sort directives using the up and down arrow buttons.

Only balance training data. If a partition field is present in the stream, this option balances data in the training partition only. In particular, this may be useful if generating adjusted propensity scores, which require an unbalanced testing or validation partition. If no partition field is present in the stream (or if multiple partition fields are specified), then this option is ignored and all data are balanced.