Multi-modal metric

Work package 4

Central for Work Package #4 is 1) to develop and test the performance of the multi-modal metric incorporating data from the three WP’S: Brain, Blood and Behaviour.

 

Methodology: Data about dynamics in neural activity, hormonal and metabolite status, behaviour and subjective sensations will contribute to the multi-modal metric. We will use dimensionality reduction methods to identify the principle components within each data modality that express the largest variance. Taking the highest ranked components for each modality for each time point, we will regress these components onto the target variable of ad libitum meal energy consumption for all subjects in a random effects regression analysis. This provides modality and time-specific statistics of interest, such that the predictive value of each data modality can be evaluated at each time point. From this regression analysis we can extract the weightings of the multimodal data that best predict energy consumption.