Table 1.

Illustration of stepwise best-fitting statistical model selection procedure

ModelsLog-likelihood increaseAIC decrease
1.RT ∼ congruency + (1 | Subj)
2.… + condition30.154.2
3.… + alphaABS + congruency: alphaABS5.46.7
4.… + Target-SSEP: condition + Flankers-SSEP: condition86.3156.5
5.… + theta power61.6121.3
  • The first row represents the baseline model which includes the fixed effect of congruency and a random intercept per subject. In each subsequent row, additional fixed-effects factors and their interactions (denoted with colon sign) are gradually included. Increase in log-likelihood and decrease in AIC indicates an increase in goodness of fit obtained by adding an additional predictor compared with the simpler model in the row above. Note that all models presented in this table had only a random effect of intercept.

  • alphaABS, Distance between 10 Hz and IAF; Target-SSVEP and Flankers-SSVEP, single-trial SSVEP amplitude in the poststimulus time window (2000–2600 ms) for target and flanker stimuli, respectively.