Measurement
Measurement is the foundation of quantitative science, and it’s
especially important in psychological and social sciences. If a variable
was not measured with good precision and validity, any research results
based on it is like picking out patterns from random noise, or even
highly misleading when systematic biases are present.
Current projects
- New algorithms for adjusting parameter estimates for measurement
noninvariance (Mark Lai & Winnie Tse)
- Measurement invariance with categorical data (Winnie Tse)
- Quantifying measurement noninvariance in a meaningful way (Yichi
Zhang)
- Bias-correction methods for heterogeneous measurement errors (Mark
Lai)
- Developing a Multidimensional Psychometric Framework on the Impact
of Item Bias on Classification (Mark Lai & Yichi Zhang & Meltem
Ozcan, Contract with Army Research Institute)
- Synthesis of invariance research (Jimmy Zhang & Haley Yue)
Multilevel Modeling
Multilevel modeling, also called mixed-effect models, is an extremely
powerful framework with very broad applications for quantitative
science. It makes efficient use of data by adaptively pooling
information across clusters (i.e., schools, states, companies, persons,
or any continuous groups, including continuous ones).
Current projects
- Reliability for clustered and longitudinal data (Mark Lai)
- Bias-correction for contextual and between-level effects (Mark Lai
& Yichi Zhang)
- The Impact of Ignoring Parameter Uncertainty on Sample-Size Planning
for Cluster-Randomized and Multisite Randomized Trials (Winnie Tse &
Mark Lai, funded by Spencer Foundation)