I know that a value over 33333 in any ideal file causes that TR to be skipped during analysis. However, what would the effect of putting a very high number (for example, 99999) in a task regressor file at particular timepoints (instead of 0's or 1's) if the file is used as an orthogonal regressor.