The output dataset contains a 32-bit-real-valued image which is that linear combination
of the input model components
which most closely approximates the expectation value
of the (masked) image
. The best-fit amplitudes
and the names of the component datasets are stored in the image array header in a table named COMPDATA which has columns as follows: