Hi Aditya,

A location within the brain barely ever sees a 6 percent
signal change. To hit 200 means that location is seeing
a 100 percent signal change. So clipping values at 200
should not affect voxels that you care about.

It is done because most datasets are scaled shorts, which
give you about 3 significant digits. With a max of 200,
that gives you maybe 1 place after the decimal.

If the maximum is not capped, it can easily rise into the
thousands at some locations. That can cost you an entire
decimal place of accuracy (or more) to capture those values
that do not even matter. Meaning you would not be able to
store 102.5, it would have to be either 102 or 103. That
amounts to adding artificial noise into your data.

So it is done to preserve the accuracy of short datasets.

Note that afni_proc.py has a -scale_no_max option to turn
off the application of any limit.

- rick