Abstract: Functional near infrared spectroscopy (fNIRS) is an emerging non-invasive optical technique to monitor the cortical hemodynamic response. Generally, parametric statistical methods are used to analyze fNIRS data, requiring certain strong assumptions that may fail in fNIRS data. This paper illustrates the application of non-parametric alternatives, such as permutation and bootstrap methods, which require fewer and weaker assumptions. We demonstrate that the proposed methods can increase the statistical significance of results when compared to the equivalent parametric methods in controlling familywise error rate in fNIRS group studies.
Key words and phrases: Adjusted p-values, maximum t correction, multiple comparison, multiple testing problem, non-parametric test, exchangeability, Type I error, optical imaging.