Mithra,
> I am using the diagnostic toolbox SPMd2. In the paper "Diagnosis and
> exploration of massively univariate neurimaging models" by Luo and
> Nichols,they talk about a plot "Temporal Outlier count" in "Scan
> summary". This paper was based on SPM99 and SPMd99. The version I
> am using is SPM2 and SPMd2. This version gives "Temporal Outliers
> proportion" instead of "Outliers count" in the "Scan summary". Can
> anyone please tell me the connection between outlier proportion and
> outlier count. Also, is it possible for me to view the outlier count
> in this version.
We changed the outlier plot because it was hard to interpret
the raw outlier count. E.g., you would expect more outliers by chance
with a 100,000 voxel full high-res brain than with a 4,000 4-slice
image. So we changed it to outlier proportion
The temporal outlier proportion is the ratio of outlier count to the
number of outliers expected under the null hypothesis of Normal data
and a well-fitting model. We define an outlier as an observation that
is greater than 3 sd from the fit. If the data are Normal and the
model fits, there should be 0.003*S outliers (0.003=P(|Z|>3)), where S
is the number of brain voxels. So if C is the outlier count, the plot
shows C/(0.003*S) (to be precise, it's not 0.003*S, but rather
spm_Ncdf(-3)*S).
[Yikes! I just checked the code, and discovered a bug! Instead of
spm_Ncdf(-3) we have spm_Ncdf(3). So sorry! Because of the outlier
proportion has been off by about 3 digits! The relative heights on
the plots are still useful, but the y-axis isn't right.
The fix is to use this revised spmd_comp_SS.m file
http://www.sph.umich.edu/ni-stat/SPMd/fix/spmd_comp_SS.m
]
If you prefer seeing raw outlier counts, then edit spmd_SS.m around
line 240, changing the line
D(3).ydata = Toutl.prop; %-change to proportion
to instead read
D(3).ydata = Toutl.ts; %-Absolute outlier count
and then change the lines below specifying the title and help
description.
Thanks for you question!
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|