Dear Christopher, Unfortunately, still does not work: For example two samples are: a=[1 2 1 2 1 2]; b=[10 11 10 11 10 11]; c=[a,b] % pooled samle std(c) =4.7290 % standard deviation of pooled sample std(a) =0.5477 % standard deviation of separate samples std(b) =0.5477 sqrt( (5*(std(a)^2)+5*(std(b)^2) )/ 10) % "total" standard deviation =0.5477 0.5477 is obviously less than 4.7290 g=sqrt((6*( std(a)^2+mean(a)^2) + 6*(std(b)^2 + mean(b)^2) )/12 - mean(c)^2) g= 4.5332 Close, but not the same. Maybe there is a mistake in the code... Could you please have a look at the formula I used? Thank you very much for you help! Sincerely yours, Vladimir >________________________________ > From: "Watson, Christopher" <[log in to unmask]> >To: Vladimir Bogdanov <[log in to unmask]>; "[log in to unmask]" <[log in to unmask]> >Sent: Tuesday, March 18, 2014 4:00 PM >Subject: RE: [SPM] simple question of statistic, how to calculate total variance? > > >You can try this: http://stats.stackexchange.com/a/43183 >________________________________________ > >From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Vladimir Bogdanov [[log in to unmask]] >Sent: Tuesday, March 18, 2014 6:55 AM >To: [log in to unmask] >Subject: [SPM] simple question of statistic, how to calculate total variance? > >Dear SPM experts, > >I have a very basic question, not specific to neuroimaging. > >I have statistical parameters for few samples: M1, M2, M3..., SD1, SD2, SD3..., N1, N2, N3... (mean, standard deviation and the size of each sample). I dont have original data. I want to estimate total M and SD. Estimation of total mean seems easy: (M1*N1+M2*N2+M3*N3... ) N1+N2+N3... > >However I have a problem with total SD. With a model data sample the formula: sqrt( ((N1-1)*SD1^2+(N2-1)*SD2^2+(N3-1)*SD3^2) / (N1+N2+N3-3) ) does not work. Perhaps the problem is that valiance between meas (M1 M2 and M3) is not taken into account. > >Thank you very much for nay help! > >Sincerely yours, >Vladimir > >