search

treder / MVPA-Light

Matlab toolbox for classification and regression of multi-dimensional data
MIT License
67 stars 34 forks source link

SEM vs. SD #33

Open nastarandarjani opened 2 years ago

nastarandarjani commented 2 years ago

I wonder why standard deviation in line 698 of examples/understanding_statistics.m didn't divide to the square root of population. I think, it is supposed to be standard error of means, due to the mean estimation of within groups.

treder commented 2 years ago

Is this a general question about the computation of perf_std? The line you mention isn't specifically about SEM/SD. More generally, when showing errobars either SEM or SD are valid depending on what you want to show. I like SD because it's constant irrespective of N whereas SEM scales down with sqrt(N), so you can make it arbitrarily small by increasing N. Nothing wrong with computing SEM though and showing that, I usually prefer SEM for visualisation in papers but I've seen both in the literature.

nastarandarjani commented 2 years ago

Isee, thank you.

On Sat, May 21, 2022, 9:55 PM Matthias Treder @.***> wrote:

Is this a general question about the computation of perf_std? The line you mention isn't specifically about SEM/SD. More generally, when showing errobars either SEM or SD are valid depending on what you want to show. I like SD because it doesn't scale with N, it's constant irrespective of N whereas SEM scales down with sqrt(N), so you can make it arbitrarily small by increasing N. Nothing wrong with computing SEM though and showing that.

— Reply to this email directly, view it on GitHub https://github.com/treder/MVPA-Light/issues/33#issuecomment-1133690622, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQHHE7437A6BZMJJ72NIPI3VLEMCHANCNFSM5WSBZOMA . You are receiving this because you authored the thread.Message ID: @.***>