Novel Approach for Analyzing Neurochemical Data: Generalized Linear Models

Authors: Ada Hepner, Deborah Kreiss

SUNY Campus: Binghamton University

Presentation Type: Poster

Location: Old Union Hall

Presentation #: 53

Timeslot: Session A 9:00-10:00 AM

Abstract: Published data concerning measurement of neurochemical levels is typically analyzed using analysis of variance (ANOVA). However, most analyses are performed without verifying proper statistical assumptions of linearity, independence, normality, and homoscedasticity. Application of ANOVAs to data that violates the required assumptions conflicts with existing statistical protocols and may lead to misinterpretation of results. The aim of this study was to employ generalized linear models to determine the effects of sudden removal of high social enrichment on neurochemical levels in rats. In the social enrichment study, both male (n=28) and female (n=28) subjects were exposed to five weeks of high social enrichment, which included frequent handling sessions and ‘playdates’ with 13 other rats. For the four weeks following, experimental rats were deprived of these social interactions, while the control rats continued to experience high social enrichment. Neurochemical concentrations in post-mortem tissue from the medial and lateral thalamic nuclei were measured using High Performance Liquid Chromatography (HPLC). The chemicals measured included the neurotransmitters dopamine, serotonin, and norepinephrine, as well as the metabolites DOPAC and 5-HIAA. To assess this data set using generalized linear regression models, data were clustered by neurotransmitters and evaluated for the proper statistical assumptions. Then, data were transformed using the Box-Cox method of power transformations in order to better fit the data to a linear model. Finally, data were analyzed using the appropriate linear model. Application of generalized linear models enhances the ability to correctly interpret neurochemical findings and to derive appropriate conclusions.