Investigation of the multiple comparisons problem in the analysis of the wave train electrical activity of muscles in Parkinson’s disease patients

Abstract

Abstract A new method has been developed for the analysis of the wave train electrical activity of muscles based on the wavelet analysis and ROC analysis that enables to study the time-frequency characteristics of electromyograms (EMG) and acceleration (ACC) signals in patients with Parkinson’s disease (PD). The idea of the method is to find local maxima (that correspond to the wave trains) in the wavelet spectrogram and to calculate various characteristics describing these maxima: the leading frequency, the duration of the wave trains in periods, the bandwidth of the wave trains, the number of wave trains per second. The degree of difference between a group of patients and a control group of volunteers in the space of these parameters is analyzed. ROC analysis is used for this purpose. The functional dependence of AUC (the area under the ROC curve) on the values of the boundaries of parameters’ ranges under consideration is investigated. The developed method involves investigation of a big number of ranges of selected characteristics; therefore a multiple comparisons problem appears during statistical hypothesis testing. It is necessary to find a compromise between the degree of detail of the studied characteristics and the magnitude of the Bonferroni correction. The paper describes the statistical hypothesis testing on the data of early Parkinson’s disease patients.