Hi,

There is no such thing as a 16-bit floating point type in NIFTI (or in standard C/C++).
Hence if you want to store non-integers (which is the most common case) then the smallest option is 32-bit float which is generally what FSL uses.

All the best,
Mark


On 10 Aug 2013, at 21:06, Keith Jamison <[log in to unmask]> wrote:

Thanks for the suggestion.  Why would I run into problems?  I'm mostly trying to reduce file size and it seems like a 16bit floating point should be more than sufficient for most analysis.  What sort of scenarios would require greater precision?

-Keith


On Sat, Aug 10, 2013 at 2:58 PM, wolf zinke <[log in to unmask]> wrote:
Uh, sorry, just realized that I misread your email (subject got cropped), and that you have not been asking for a different output type, but a fixed data type. Here I do not know how to change it globally. Only script wise you can use a variable like DT="-odt float" and append this after each fslmaths call.

However, I guess forcing data types might lead in some cases to problems, so I would be careful to do taht.


wolf

On 10/08/13 21:27, Keith Jamison wrote:
Is there a way to specify the data format that fsl will output by default?  I know I can convert them manually after each command but I would prefer to just set it once globally.

-Keith