3 ms·It’s not because of analytical convenience, it’s because of the central limit theorem.by xadhominemx 1y agoIt’s not because of analytical convenience, it’s because of the central limit theorem.jampekka 1y agoNot everything is a linear combination of large number of (IID) samples, and thus not everything is gaussian distributed.kgwgk 1y agoYou’re implying that many things are though.jampekka 1y agoYes, and I was explicit about it in another comment in this post.kgwgk 1y agoOk, so we all agree that in most cases the reason to minimize square error is that it’s the appropriate thing to minimize - not that it has an analytical solution, convenience or tradition.