Permutation testing in analyzing the changes in the intensity of hypoglycaemic episodes occuring within 24 hours from previous episodes

Abstract

This study is motivated by the issue that arises when modelling the consistency of symptoms reporting by individual patients during hypoglycaemic episodes. It is argued that any episode occurring within 24 hours from a previous episode may reduce the intensity of subsequent hypoglycaemic episodes. Such episodes were excluded from further consideration because they could potentially affect our model given that it is based on the patient’s inherent propensity to report a given symptom and the intensity of that symptom in a given episode. The aim of this study is to investigate the effect of adding data from episodes that occurred within 24 hours from their preceding episode. We develop a permutation testing to investigate whether episodes’ intensity exhibits change that goes beyond random variation when episodes are in temporal proximity. Further analysis consists of comparison between the two cases i.e. excluding episodes within 24 hours and including those episodes with regards to their correlations between the intensity of the episodes. The analysis shows that adding these new episodes does not have a significant impact on the intensity of episodes.This study is motivated by the issue that arises when modelling the consistency of symptoms reporting by individual patients during hypoglycaemic episodes. It is argued that any episode occurring within 24 hours from a previous episode may reduce the intensity of subsequent hypoglycaemic episodes. Such episodes were excluded from further consideration because they could potentially affect our model given that it is based on the patient’s inherent propensity to report a given symptom and the intensity of that symptom in a given episode. The aim of this study is to investigate the effect of adding data from episodes that occurred within 24 hours from their preceding episode. We develop a permutation testing to investigate whether episodes’ intensity exhibits change that goes beyond random variation when episodes are in temporal proximity. Further analysis consists of comparison between the two cases i.e. excluding episodes within 24 hours and including those episodes with regards to their correlations betwe...