Data from: James-Stein estimator improves accuracy and sample efficiency in human kinematic and metabolic data
Data files
Nov 19, 2024 version files 64.81 KB
-
README.md
4.40 KB
-
Three_tasks_parameter_estimates_and_standard_error_lists.zip
60.40 KB
Abstract
In the associated manuscript, we show that James-Stein estimator helps improve statistical estimates for kinematic and metabolic problems by pooling data from multiple subjects. We considered three types of estimation problems: foot placement control, metabolic cost of walking in circles and metabolic cost of resting. In this dataset, we have shared the individual statistical estimates for each subject for each type of data, for each trial duration, along with their standard errors. The data in MATLAB's .mat format and can be opened with MATLAB, or free software such as octave and python.
https://doi.org/10.5061/dryad.3j9kd51v9
Description of the data and file structure
The repository has three subfolders, one for each type of data and estimation problem.
1) foot placement control - the sensitivity of foot placement to sideways perturbations
2) resting metabolics - the metabolic cost of sitting/resting
3) walking metabolics - the metabolic cost of walking in a circle
Both the statistical estimates and their standard errors are provided for all the subjects, all these different estimation problems, and for different trial durations, as provided in the associated manuscript. The file naming makes it clear what duration each dataset is for and what quantity is being stored in the file.
Software: All data files are .mat files, namely MATLAB data files, and can be loaded with the ‘load’ command in MATLAB. Free software such as Python and Octave have appropriate corresponding commands for opening such files.
We now describe the variables and the format for the data in each folder in more detail.
Foot placement control
The file names in this folder all have a name of the form ‘xFootByxHip_LtoR_N_Steps.mat’, where N ranges from 15 to 100 in steps of 5. e.g., xFootByxHip_LtoR_15_Steps.mat, xFootByxHip_LtoR_20_Steps.mat, etc.
The files in this folder contains the sensitivity of foot placement relative to hip (pelvis) state. As the file names above indicate, the quantity reported is the sensitivity of xFoot (foot position in sideways direction) to xHip (hip position relative to previous stance foot in the sideways direction). The LtoR in the file name refers to the fact that the mappings are from hip position during left stance (L) to foot position during the next right stance (R). The number N refers to the number of left to right transitions used to infer the sensitivity.
Each file can be loaded in MATLAB using the command, where N = 15, … 100:
load(‘xFootByxHip_LtoR_N_Steps.mat’)
Two 8 x 1 array variables will be loaded into the MATLAB workspace.
xFootByxHip_LtoR_Estimate - the value of the sensitivity of xFoot to xHip for 8 human participants
xFootByxHip_LtoR_SE - the standard error of the xFootByxHip_LtoR_Estimate for the 8 subjects
The motion capture data was obtained using Vicon T20 system with Nexus 1.x.
Resting metabolics
The file names in this folder all have a name of the form ‘metabolicCost_resting_N_points’, where N ranges from 5 to 40 in steps of 5. e.g., metabolicCost_resting_5_points.mat, metabolicCost_resting_10_points.mat, etc.
The files in this folder contains the mean resting metabolic rate for 27 subjects. The number N refers to the number of individual metabolic rate samples used to infer the average.
Each file can be loaded in MATLAB using the command, where N = 5, … 40:
load(‘metabolicCost_resting_N_points.mat’)
Two 27 x 1 array variables will be loaded into the MATLAB workspace.
metCost_Estimate - the value of the mean metabolic rate for 27 human participants
metCost_SE - the standard error of the metCost_Estimate for the 27 subjects
The metabolic data was obtained using Oxycon Mobile indirect calorimetry system.
Walking metabolics
The file names in this folder all have a name of the form ‘metabolicCostWithExponentialFit_CircleWalking_N_points.mat’, where N ranges from 3 to 54 in steps of 3. e.g., metabolicCostWithExponentialFit_CircleWalking_3_points.mat, metabolicCostWithExponentialFit_CircleWalking_6_points.mat, etc.
The files in this folder contains the estimated steady state metabolic rate when the participant is walking in a circle. The metabolic data was obtained using Oxycon Mobile indirect calorimetry system. The number N refers to the number of metabolic rate samples used to fit an exponential to estimate the steady state value of the metabolic rates.
Each file can be loaded in MATLAB using the command, where N = 3, … 54:
load(‘metabolicCostWithExponentialFit_CircleWalking_N_points.mat’)
Two 11 x 1 array variables will be loaded into the MATLAB workspace.
metCost_Estimate - the value of the steady state metabolic rate for 11 human participants
metCost_SE - the standard error of the metCost_Estimate for the 11 subjects
The estimates and standard errors were computed as follows:
- Foot placement control sensitivities was estimated from linear regression over motion capture data, using methods identical to those in Yang and Srinivasan [1], Perry and Srinivasan [2], Joshi and Srinivasan [3], and Seethapathi and Srinivasan [4]. The estimates and standard errors were obtained from the linear regression software fitlm in MATLAB.
- The resting metabolic data just involves simple averages and are from Hanford and Srinivasan [5], Seethapathi and Srinivasan [6], and Brown, Seethapathi, and Srinivasan [7]. The estimates and standard errors were obtained by elementary formulas for mean and standard error.
- The exponential fit to the walking metabolic rate is performed using fminunc in MATLAB to minimize a mean squared error between an exponential a0 + a1*exp(-lambda t) and the data [7].
References
[1] Wang, Yang, and Manoj Srinivasan. "Stepping in the direction of the fall: the next foot placement can be predicted from current upper body state in steady-state walking." Biology letters 10, no. 9 (2014): 20140405.
[2] Perry, Jennifer A., and Manoj Srinivasan. "Walking with wider steps changes foot placement control, increases kinematic variability and does not improve linear stability." Royal Society open science 4, no. 9 (2017): 160627.
[3] Joshi, Varun, and Manoj Srinivasan. "A controller for walking derived from how humans recover from perturbations." Journal of The Royal Society Interface 16, no. 157 (2019): 20190027.
[4] Seethapathi, Nidhi, and Manoj Srinivasan. "Step-to-step variations in human running reveal how humans run without falling." Elife 8 (2019): e38371.
[5] Handford, Matthew L., and Manoj Srinivasan. "Sideways walking: preferred is slow, slow is optimal, and optimal is expensive." Biology letters 10, no. 1 (2014): 20131006.
[6] Seethapathi, Nidhi, and Manoj Srinivasan. "The metabolic cost of changing walking speeds is significant, implies lower optimal speeds for shorter distances, and increases daily energy estimates." Biology letters 11, no. 9 (2015): 20150486.
[7] Brown, Geoffrey L., Nidhi Seethapathi, and Manoj Srinivasan. "A unified energy-optimality criterion predicts human navigation paths and speeds." Proceedings of the National Academy of Sciences 118, no. 29 (2021): e2020327118.