Difference between revisions of "Lab Manual:MSSD"
From SpanLabWiki
| (2 intermediate revisions by 2 users not shown) | |||
| Line 3: | Line 3: | ||
I wrote a script to calculate MSSD in AFNI. | I wrote a script to calculate MSSD in AFNI. | ||
| − | + | For those who missed Greg's discussion about MSSD: MSSD is calculated like variance, but instead of the average square of difference from the mean, it is the average square of differences from the previous time point. | |
| + | The script can take some time to run, especially if you have 54 subjects, a 34 minute scan, and have resampled to teeny voxels. | ||
| + | This script is useful if you want to do a whole brain analysis of MSSD across subjects. DJY 20081022 | ||
| + | |||
| + | <font size = 3> | ||
| + | <source lang="bash"> | ||
#! /bin/csh | #! /bin/csh | ||
| Line 56: | Line 61: | ||
rm -f prevtotalsum_${output}* | rm -f prevtotalsum_${output}* | ||
| + | </source> | ||
| + | </font> | ||
Latest revision as of 16:53, 5 June 2009
MSSD is a calculation of noise similar to variance, however instead of variance from the mean, the variance is calculated from the previous time point.
I wrote a script to calculate MSSD in AFNI.
For those who missed Greg's discussion about MSSD: MSSD is calculated like variance, but instead of the average square of difference from the mean, it is the average square of differences from the previous time point.
The script can take some time to run, especially if you have 54 subjects, a 34 minute scan, and have resampled to teeny voxels.
This script is useful if you want to do a whole brain analysis of MSSD across subjects. DJY 20081022
<source lang="bash">
#! /bin/csh
# created by djy 20081021 for FINRA
# this script will calcualte the MSSD for a block of data
#DEFINE VARIABLES HERE
set data = normfBIAS+tlrc # this is your data file
set output = bias_MSSD # this is the prefix of the output data
@ startTR = 0 # set this to the starting TR, usually 0 if you want to calculate MSSD from the very beginning
@ TRcount = 1010 # set this to the total TRs, if you want to model the entire block
@ endTR = $TRcount - 1 #set this to the TR you want to end on.
# The real meat starts here.
@ TRsum = $endTR - $startTR
@ currentTR = $startTR + 1 # set the current TR for the loop to the start TR + 1.
@ previousTR = $currentTR - 1
echo $currentTR # print which TR we are starting with
3dcalc -float -a $data'['$previousTR']' -b $data'['$currentTR']' -expr "(b-a) * (b-a)" -prefix 'tempsum_'$output
3dcalc -float -a 'tempsum_'$output+tlrc -expr 'a' -prefix 'prevtotalsum_'$output
rm -f tempsum_${output}*
@ currentTR = $currentTR + 1
while ( $currentTR < $endTR + 1)
echo $currentTR
@ previousTR = $currentTR - 1
3dcalc -float -a $data'['${previousTR}']' -b $data'['$currentTR']' -expr '(b-a) * (b-a)' -prefix 'tempsum_'${output}
3dcalc -float -a 'tempsum_'$output'+tlrc' -b 'prevtotalsum_'$output'+tlrc' -expr 'a+b' -prefix 'newtotalsum_'${output}
rm -f tempsum_${output}*
rm -f prevtotalsum_${output}*
3dcalc -float -a 'newtotalsum_'$output'+tlrc' -expr 'a' -prefix 'prevtotalsum_'$output
rm -f newtotalsum_${output}*
@ currentTR = $currentTR + 1
end
echo "total TR's for average = "$TRsum
3dcalc -float -a 'prevtotalsum_'$output'+tlrc' -expr 'a/('$TRsum')' -prefix $output
rm -f prevtotalsum_${output}*
</source>
