Table 2 Model predictive performance, representational reproducibility, and model robustness when trained to predict participant diagnosis of autism vs. NT, evaluated according to the procedure for determining model robustness. 10 repetitions of disjoint data splits were used to quantify model robustness metrics. Results achieved with model parameters corresponding to the highest model robustness score for a given modeling pipeline. Standard error (SE) reported.

From: Model selection to achieve reproducible associations between resting state EEG features and autism

Model

Predictive Performance (SE)

Representational Reproducibility (SE)

Model Robustness (SE)

Logistic Regression

0.693 ± 0.004

0.542 ± 0.019

0.210 ± 0.010

Logistic Regression + SIRT

0.700 ± 0.004

0.678 ± 0.016

0.272 ± 0.010

  1. Note. Abbreviations: SIRT, Scale-Invariant Residuals Transform; AUROC, area under the receiver operating characteristic curve; SE, standard error.
Back to article page