Population Stability Index (PSI)
PSI is a measure of the degree to which the distribution of a population has shifted over time or between two different samples of a population in a single number.
Population Stability Index (PSI)
The population stability between two population samples is measured with the use of PSI score. When the distributions are reasonably stable, it is a useful metric for both numerical and categorical data.
Equation: PSI = (P - Q)ln(P/Q)
According to the ln(P/Q) term, a significant change in a bin that represents a small percentage of the distribution will have a greater impact on PSI than a large change in a bin that represents a significant portion of the distribution.
It is important to remember that the population stability index only reflects changes in the population distribution of the data. However, performance may or may not suffer as a result of this. Therefore, you may use PSI to confirm that the population distributions have truly changed when you see a performance drop. You can also include PSI checks in your monitoring strategy.
PSI closer to '0' means that the distribution is stable. 0.1 is used as the default threshold in AryaXAI.
Liked the content? you'll love our emails!
See how AryaXAI improves
ML Observability
Learn how to bring transparency & suitability to your AI Solutions, Explore relevant use cases for your team, and Get pricing information for XAI products.
