Information value

1. Information value

So you now know how to derive new variables based on your domain knowledge. But do you think these new variables along with the already existing ones can predict turnover? To examine how each variable influences turnover you can use a technique called Information value.

2. Understanding Information value

Information value is a data exploration technique that helps determine which independent variable in a dataset has predictive power or influence on the dependent variable. Simply speaking, Information value provides a measure of how well independent variables such as employee tenure, level, compa-ratio etc. are able to distinguish between a binary response, i.e., "Active" versus "Inactive". If a variable has low information value, it may not do a sufficient job of explaining the dependent variable, in our case, turnover.

3. Calculating Information value

The Information value is calculated using the formula shown on slide here. Here, events refer to inactive employees, i.e., people who have left the organization, whereas non-events refer to active employees. The detailed mathematical explanation is beyond the scope of this course and we will be using the Information package to calculate information value.

4. Calculating Information value

The create_infotables() function from the Information package calculates the information value for all the variables in the dataset. This function requires the dataset as its first argument followed by the target variable, i.e., the variable we want to predict, in this case, turnover. The Summary element in the result contains the information value of all the variables in the dataset. Here, the IV of percent hike variable is 1.14 whereas IV of compa_ratio variable is 0.29. Let's understand how to interpret these values.

5. Information value (IV) table

We can interpret the predictive power of the variables using this table. percent_hike which has IV of 1.14 is a strong predictor, whereas compa_ratio with an IV of 0.29 is a moderate predictor of turnover.

6. Let's practice!

Go ahead and calculate information value of the variables in your dataset.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.