Multilevel Modeling of Cognitive Function in Schizophrenic Patients and Their First Degree Relatives.

Area of Science:

• Psychiatry
• Biostatistics
• Cognitive Neuroscience

Background:

• Schizophrenia affects cognitive function, impacting patients and their families.
• Understanding cognitive performance variations within families is crucial.

Purpose of the Study:

• Compare mean cognitive performance across schizophrenia patients, healthy relatives, and controls.
• Investigate differences in cognitive performance variances between these groups.
• Address methodological challenges in multilevel modeling for nested family data.

Main Methods:

• Applied multilevel modeling to analyze cognitive function data from subjects nested within families with repeated measures.
• Utilized non-parametric maximum likelihood (NPML) for non-normal random effects distributions.
• Employed generalized linear mixed models for skewed response data.
• Compared penalized quasilikelihood (PQL) with Gaussian quadrature for binomial responses.

Main Results:

• Non-parametric maximum likelihood (NPML) yielded different conclusions compared to conventional multilevel models for specific hypotheses.
• Allowing between-subject variance to differ between groups provided nuanced insights.
• Generalized linear mixed models and Gaussian quadrature demonstrated varying accuracy for skewed and binomial data.

Conclusions:

• Methodological choices in multilevel modeling significantly influence findings in schizophrenia cognitive research.
• Advanced statistical techniques are essential for accurately modeling complex cognitive data in psychiatric populations.
• Further research should explore the implications of these statistical approaches for clinical understanding.