Modelling punctuated similarity

Abstract

Inter-subject, pairwise similarity models provide a methodological resource for flexibly measuring complex, non-linear relationships between brain and behavior. Similarity models, however, can extend beyond brain behavior relationships and can be readily applied to any data where they may be useful. The work presented in this paper introduces a new way of modelling similarity, termed punctuated similarity, where dissimilarity is modelled at a specific point within a given sample of continuous data. With this model researchers can select a value at which they expect dissimilarity or similarity to occur and evaluate it against real similarity distributions. To demonstrate how this model works two separate proof-of-concept experimental analyses were conducted on publicly available data, the results of which are reported in this paper. The first investigated puberty as a critical time-point for growth in a sample of children aged five to nineteen obtained from the NCD Risk Factor Collaboration. The second investigated the COVID-19 pandemic as a critical time-point for market volatility in stock price data for market indices from seven different countries. Both experiments showed that the similarity model presented in this paper was effective at modelling critical time points for increased variance in stock market prices and growth development of children. Finally, this paper is accompanied by an open-source R library to recreate the similarity models presented, providing a tool for future researchers to use in their own analysis.