Measurement bias occurs when the data collected is faulty measured and when it differs from the real-world data.
Example: When you record the audio to train an AI model with one microphone and then use another microphone during the production.
Example: When a quality inspector defines the defective products as non-defective and uses that data to train a model.
How to prevent measurement bias
To prevent measurement bias, you can check:
If the measurement tool is working properly
If there is an inconsistency between measurements with several tools
If the data collected is faulty by human mistake