Hi Christine,
I do not have lehuakona in my desktop now, but while waiting for Jamie
to try it on, just a quick answer that might help.
It seems that this problem happen because of the method=global, not
due to the order of fitting, I think. We tried
mfittrend in=${in_object} axis=3 order=1 out=temp_01 auto
mask=temp_mask1 method=single variance=false subtract=false 'clip=[2,3]'
and it seems OK. But
mfittrend in=${in_object} axis=3 order=1 out=temp_02 auto
mask=temp_mask2 method=global variance=false subtract=false
'clip=[1.5,2.5,2,3,5]
is different.
I think a quick way just to look at the output from this mfittrend, in
above example temp_02, the affected regions seems to appear
surprisingly high or lower than the rest of the pixel, which seems to
be fitting OK.
Hope that help a little bit.
Cheers,
-Boon-
On 24 Sep 2009, at 14:08, Christine Wilson wrote:
> Hi, Jamie,
>
> Thanks for bringing this to our attention.
>
> From the release notes, it looks like Lehuakona was active from
> 2008-11-12 to 2009-07-27. So this means that all the work that my
> student did this summer was probably done on lehuakona, since I
> don't think we did an upgrade late in the summer.
>
> If the work has to be redone, it is a big job. I don't have time to
> work on this at the moment. So two questions:
>
> (1) does the problem ONLY occur for a first order baseline or would
> it also occur for a third order baseline? If you could test that
> quickly on your data set, that would be great.
>
> (2) is there any quick way to judge if the data are affected by this
> problem?
>
> Thanks,
> Chris
>
|