Bootstrap Prediction and Confidence Bands for Frequency Response Functions in Posturography

Vittorio Lippi · 2025

Read Bootstrap Prediction and Confidence Bands for Frequency Response Functions in Posturography by Vittorio Lippi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-20588.txt.

Introduction

Vittorio Lippi proposes a better way, in a 2025 preprint, to judge the results of balance experiments. In posturography, a person stands on a platform while a stimulus is applied. The outcome is summed up by a frequency response function, which links the stimulus to body sway at each frequency. That function is a bundle of complex values, one per frequency. Testing each value on its own, with standard corrections, is overly cautious.

Averaging the values instead forces an arbitrary choice on the researcher. Lippi offers bootstrap prediction and confidence bands as a general fix. The bootstrap resamples the data to show the range of curves the experiment could plausibly produce. Bands around the whole curve keep the frequencies tied together. The method aims to be fair without being needlessly strict. It applies to posture control work in general, not one setup alone.

Catalog record

CREATORVittorio Lippi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-20588

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Bootstrap Prediction and Confidence Bands for Frequency Response Functions in Posturography Vittorio Lippi Abstract The frequency response function (FRF) is an established way to describe the outcome of experiments in posture control literature. The FRF is an empirical transfer function between an input stimulus and the induced body segment sway profile, represented as a vector of complex values associated with a vector of frequencies. For this reason, testing the components of the FRF independently with Bonferroni correction can result in a too-conservative approach. Performing statistics on scalar values defined on the FRF, e.g., comparing the averages, implies an arbitrary decision by the experimenter. This work proposes bootstrap prediction and confidence bands as general methods to evaluate the outcome of posture control experiments, overcoming the foretold limitations of previously used approaches. MSC Classification]62G09, 92C55, 65C60,62G15,93C05,62M15,92C50 keywords Frequency response function, Bootstrap, confidence bands pacs [ 1 Introduction The frequency response function (FRF) is a usual way to describe the outcome of experiments in posture control literature. In particular, the FRF is an empirical transfer function between an input stimulus and the induced body movement with an additional step of averaging the value at some sets of frequencies, as explained in detail in section 2.1 . By definition, the FRF is a complex function of frequency. When statistical analysis is performed to assess differences between groups of FRFs (e.g., obtained under different conditions or from a group of patients and a control group), the FRF’s structure should be considered. Usually, the statistics are performed by defining a scalar variable to be studied, such as the norm of the difference between FRFs, or considering the components independently that can be applied to real and complex components separately Lippi et al. (2023) ; Akçay et al. (2021) , in some cases both approaches are integrated, e.g., the comparison frequency-by-frequency is used as a post-hoc test when the null hypothesis is rejected on the scalar value Lippi et al. (2020) . The two components of the complex values can be tested with multivariate methods such as Hotelling’s T2 as done in Assländer and Peterka (2014) on the averages of the FRF over all the frequencies, where a further post hoc test is performed applying bootstrap on magnitude and phase separately. The problem with the definition of a scalar variable as the norm of the differences or the difference of the averages in the previous examples is that it introduces an arbitrary metric that, although reasonable, has no substantial connection with the experiment unless the scalar value is assumed a priori as the object of the study as in Lippi et al. (2020) where a human-likeness score for humanoid robots is defined on the basis of FRFs difference. On the other hand, testing frequencies (and components) separately does not consider that the FRF’s values are not independent and applying corrections for multiple comparisons, e.g., Bonferroni, can result in a too-conservative approach destroying the power of the experiment. In order to properly consider the nature of the FRF, a method oriented to complex functions should be used. In Lippi (2023) a preliminary method based on random field theory inspired by Pataky et al. (2016) was presented: to take into account the two components (imaginary and real) as two independent variables, the fact that the same subject repeated the test in the two conditions, a 1-D implementation of the Hotelling T2 is used as presented in Pataky et al. (2014) but applied in the frequency domain instead of the time domain. Several approaches to statistical analysis of continuous data based on confidence bands are reviewed in Joch et al. (2019) . In particular, the function based resampling technique (FBRT) is based on fitting data into mathematical functions and using its coefficients to estimate a variability function. A scaling factor is used to adjust the prediction coverage. The functions used are usually sinusoid; hence, the parameters are the coefficients of Fourier expansions. The limitation of FBRT is that the choice of the functions (i.e., of the frequencies) is not always obvious: while in human movement signals such as joint trajectories during gait assuming a low-pass power-spectrum for the samples makes it reasonable to truncate its Fourier expansion, i.e. take the first N terms Lenhoff et al. (1999) , other biological signals like EEG may not clearly exhibit a set of frequencies that are representative in all the samples Joch et al. (2019) . In the specific case of FRFs from posture control experiments, the sample itself, by definition, provides a predefined set of frequencies (as it will be clear in § 2.1 ). This suggests that all the FRFs in the samples can be conveniently transformed to time domain signals that consistently represent all the components of the FRFs. On this basis, the present work proposes a bootstrap method for computing prediction and confidence bands. The method is implemented by transforming the FRF in the time domain, obtaining a pseudo-impulse-response (PIR) function on which such bands can be defined and computed. The results of the tests in the frequency domain are then presented and discussed. Prediction band represents the region that contains with a given probability \alpha\% a new PIR drawn from the same population of the sample under analysis. This can be used to classify a new observation as belonging or not to the sample population on which the band is computed. Confidence band for the population’s mean PIR can also be defined with a probability \alpha\% . It can be used to test the effect of a condition or the difference between groups, e.g., the null hypothesis for an experiment is that the mean difference between the PPR under two conditions is zero. This paper proposes the bootstrap method to compute the regions mentioned above following the one proposed for trajectories in Lenhoff et al. (1999) but adapted to the specific case of FRF. In the Methods section, a definition of the FRF used in posture control literature is provided for reference, and then the Bootstrap method is described. Then, a way to visualize the regions in the complex space is proposed, and finally, tests to evaluate the method are presented, using real human data from posture control experiments. The Results section presents the results of the tests. In the Discussion , the results are analyzed with emphasis on the implications for posturography and potentially for evaluating humanoid robots’ posture control. In the last section, The Code , there are links to the repository to download the Matlab MATLAB (2019) source and instructions to run the functions computing the bands and running tests. 2 Methods 2.1 The FRF The frequency response function, FRF, is an empirically computed transfer function defined between an input and output. In posture control experiments, the input is a stimulus, such as a support surface tilt with a specific profile, and the output is the body sway or the sway of a body segment. In this work, the FRF is defined as proposed in Peterka (2002) , in detail: 1. The input has a pseudo-random ternary signal profile, PRTS Davies (1970) , composed of segments at constant speed (0,+s,-s) where s is a fixed speed value usually set to get a specific amplitude of

Read the complete text