Dear stats wizard on the list,
when plotting the parameter estimates using the plot button, SPM plots
a bar graph with 10% confidence interval using the following formulae
(see spm_graph.m lines 179 and 265-266):
cbeta = SPM.xCon(Ic).c'*beta; % for the bars
CI = 1.6449; % = spm_invNcdf(1- 0.05);
CI = CI*sqrt(diag(SPM.xCon(Ic).c'*Bcov*SPM.xCon(Ic).c)); % for the error bar
I looked up the following formula fpr the confidence interval from one
of my stats books:
CI = Zcrit * (S/sqrt(n))
where
- Zcrit is the critical Z-values (e.g. 1.6449 as in the SPM code above)
- S is the standard deviation
- n is the sample size
If I understand the SPM code above correctly then
sqrt(diag(SPM.xCon(Ic).c'*Bcov*SPM.xCon(Ic).c)) calculates the standard
deviation, but somehow I cannot find the sqrt(n)-part in the SPM code.
I realize that dividing the standard deviation by sqrt(n) might be
inappropriate at the first level due to the many scans present in
a first level design, but wouldn't it be feasible to so at the second
level and devide the standard deviation by sqrt(n) where n is the number
of subjects in the analysis? Or am I missing another totally obvious
point here?
Thanks,
Jan
--
Jan Gläscher Neuroimage Nord
+49-40-42803-7890 (office) Dept. of Neurology, Bldg S10
+49-40-42803-9955 (fax) University Hospital Hamburg-Eppendorf
[log in to unmask] Martinistr. 52
20246 Hamburg
Germany
http://www.uke.uni-hamburg.de/zentren/neuro/neurologie/mitarbeiter/glaescher_jan.html
--------------------------------------------------------------------------------------
GnuPG/PGP key id: FEC4B55C
fingerprint: 5A36 1EF6 8472 117E 805A F240 3146 A410 FEC4 B55C
--------------------------------------------------------------------------------------
|