Related Experiment Videos
Problems with using sum scores for estimating variance components: contamination and measurement noninvariance.
Michael C Neale1, Gitta Lubke, Steven H Aggen
1Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, Virginia, USA 23219-1534, USA. mcneale@vcu.edu
Summary
Measurement invariance is crucial in twin studies. Without it, genetic and environmental variance estimates can be biased, especially when analyzing sum scores from multiple items.
Area of Science:
- Behavioral genetics
- Psychiatric genetics
- Quantitative genetics
Background:
- Twin studies commonly use sum scores for complex traits.
- Univariate analysis of sum scores assumes measurement invariance across zygosity.
Purpose of the Study:
- To demonstrate how absence of measurement invariance biases genetic and environmental variance estimates in twin studies.
- To highlight the impact of zygosity-dependent measurement bias on latent factor variance partitioning.
Main Methods:
- Theoretical analysis of measurement invariance in twin studies.
- Quantification of bias in genetic and environmental variance components due to sum score analysis.
- Comparison of sum score correlations with true latent trait correlations.
Main Results:
- Absence of measurement invariance across zygosity significantly biases estimates of genetic and environmental variance.
- Sum score analysis distorts both monozygotic (MZ) and dizygotic (DZ) twin correlations.
- Bias is particularly pronounced for sum scores derived from binary items.
Conclusions:
- Measurement invariance testing across zygosity is essential before analyzing sum or scale scores in twin studies.
- Multivariate genetic analysis at the item or symptom level offers a solution to mitigate bias.
- Careful consideration of measurement properties is needed to ensure valid genetic and environmental inferences from twin data.