Hi,
I think we did used to retain the datatype which led to a small loss
in accuracy after
scaling (mean 10000) but used to cause some problems with scripts
later on performing
inappropriate integer arithmetic. We then decided to save everything
that really should
be a floating point value in a floating point type, regardless of the
input image type.
This is our strong recommendation as everything is more natural and
less problematic.
All the best,
Mark
On 14 Mar 2008, at 10:10, Teemu Rinne wrote:
> Hi Mark,
>
> Ok, but there must be something more than that as
> filtered_func_data from an older feat (5.63, non-intel mac) is
> datatype 4 (int)?
>
> Cheers,
>
> Teemu
>
>
>
> Mark Jenkinson kirjoitti 14.3.2008 kello 11.41:
>
>> Hi,
>>
>> The datatype is changed because it needs to represent values that
>> are not integer, due to interpolation, smoothing, filtering, etc.
>> You really must keep the type as a floating point type in order for
>> this data to be meaningful.
>>
>> All the best,
>> Mark
>>
>>
>>
>> Teemu Rinne wrote:
>>> Hi,
>>>
>>> Datatype of my input file is 4 (short int) but filtered_func_data
>>> is 16 (float). (Feat v5.92, Intel mac binaries)
>>>
>>> 1) Why does feat change the data type? Is it ok to change
>>> filtered_func_data back to short int?
>>>
>>> 2) How can I change the feat default output datatype?
>>>
>>> --
>>> Teemu Rinne
>>> Dept psychology
>>> Univ Helsinki, Finland
>>>
>>
>
|