PT  - JOURNAL ARTICLE
AU  - Joke Durnez
AU  - Ross Blair
AU  - Russell A. Poldrack
TI  - Neurodesign: Optimal experimental designs for task fMRI
AID  - 10.1101/119594
DP  - 2017 Jan 01
TA  - bioRxiv
PG  - 119594
4099  - http://biorxiv.org/content/early/2017/03/23/119594.short
4100  - http://biorxiv.org/content/early/2017/03/23/119594.full
AB  - A recent stream of alarming publications questions the validity of published neuroimaging findings. As a consequence, fMRI teams worldwide are encouraged to increase their sample sizes to reach higher power and thus increase the positive predictive value of their findings. However, an often-overlooked factor influencing power is the experimental design: by choosing the appropriate experimental design, the statistical power of a study can be increased within subjects. By optimizing the order and timing of the stimuli, power can be gained at no extra cost. To facilitate design optimization, we created a python package and web-based tool called Neurodesign to maximize the detection power or estimation efficiency within subjects, while controlling for psychological factors such as the predictability of the design. We implemented the genetic algorithm, introduced by Wager and Nichols (2003) and further improved by Kao et al. (2009), to optimize the experimental design. The toolbox allows more complex experimental setups than existing toolboxes, while the GUI provides a more user-friendly experience. The toolbox is accessible online at www.neuropowertools.org.