Kinematic and looking behaviour features of a movement imitation task

Published: 06-03-2020| Version 4 | DOI: 10.17632/fnt6jtc5np.4
Andrius Vabalas,
Emma Gowen,
Ellen Poliakoff,
Alex Casson


All features are continuous, raw (non-normalised). Missing values and outliers are replaced with class means. # of classes: 2 (1 – autistic / 0 – non-autistic) # of data samples: 44 (22-autistic / 22-non autistic) # of features: Kinematic dataset: 120 per instruction block, Looking behaviour dataset: 48 per instruction block. Feature descriptions are given in a title row in the files and fuller descriptions are given in the paper and it’s Supplementary Methods.

Steps to reproduce

Steps to reproduce: • Split the data into train and test samples (in the files column 2 shows how data was split into training and testing samples in the original paper) • Normalise the data • Perform feature selection • Apply supervised machine learning classifier for 2 classes.