“…The maximin approach has been applied in optimal design theory to construct designs which are robust against misspecification of parameters in various models, e.g., the linear mixed model (Berger & Tan, 2004;Ouwens, Tan & Berger, 2002) or exponential regression models (Dette, Martinez Lopez, Ortiz Rodriguez & Pepelyshev, 2006). The effect of autocorrelation on optimal design has been studied before in the fMRI literature by assuming fixed parameter values in the error structure or by using real fMRI noise (Aguirre & D'Esposito, 1999;Smith et al, 2007;Wager & Nichols, 2003;Zarahn & Friston, 2002;Buračas & Boynton, 2002;Burock & Dale, 2000). As a result, the found optimal designs may not be robust against misspecification of these parameters.…”