Data Anomalies on Model Integrity

Requirements Clarification & Assessment

1. Understanding the Problem:

   • The logistic regression model relies heavily on a specific variable.
   • Due to a data quality issue, some values in this variable have lost their decimal points, turning values like 100.00 into 10000.

2. Assessing the Impact:

   • Determine if the altered values are outliers or influential points.
   • Understand the role of the variable in the model and how its distortion affects model integrity.

3. Defining the Goal:

   • Ensure the model remains valid and accurate.
   • Identify and correct the data quality issue without compromising model performance.

4. Constraints & Assumptions:

   • Assume the variable's original value range is known.
   • Assume the data quality issue is limited to decimal point loss.
   • Assume access to tools for detecting influential points (e.g., Cook's distance).