Module 05
Statistical concepts and methods
Please contact the presenter for a permission to use the content from the E-Learning modules. Thank you for your understanding.
Interview with Stef van Buuren, Prof., TNO, Project Staff
Duration: 10min
1. Comparability as a Missing Data Problem
Stef van Buuren, Prof.
Duration: 23min
Summary
Data harmonisation typically involves arbitrary ad-hoc rules like “equating categories”. An alternative is to carefully formulate the assumptions that bridge different variables, and search for information about their relation. That information forms the basis of multiple imputation of the missing information, and hence is more data-driven than plain equating. A simple example with real data shows that the two approaches may lead to substantially different results.
Learning goals
- Simple equating strategies exaggerate differences between studies.
- It is better to rely on data than to make unverified assumptions.
- Multiple imputation is a general purpose tool for principled harmonisation
2. D-score for Child Development
Stef van Buuren, Prof.
Duration: 25min
Summary
Optimal growth and development are crucial for the child to live a happy, healthy and productive life. Growth is easily measured, but this is not true for development. Major comparability problems occur when we try to combine developmental data from different sources. This lecture provides the D-score as an alternative. We hope and expect that the D-score will ease the exchange and interpretation of child development data.
Learning goals
- Instruments and scales are independent concepts
- The Rasch model ties together and filters measurement made by different instrument, and produces a scale with a fixed unit
- The D-score enables quantifying children measured by different instruments onto the same scale