How to determine data is normally distributed?
in [Matlab]
|
Prev: www.voguesneakers.com Cheap Jordans,Cheap Nike Shox R4,Cheap Nikes
Next: Enabled subsystems do not disable
From: Alfian Abdul Halin on 20 Apr 2010 03:26 Hi all, I was wondering if anyone knows how to determine that a particular set of multivariate data is normally distributed? (Hope I'm using the multivariate term correctly, since I have just learnt it...) For example... I have a collection of feature vectors, where X = (x1,x2,x3)... Overall, I am making N-observations... so X1, X2, ..., XN feature vectors are generated. My assumption is that (based on some domain knowledge), a multivariate normal distribution will result. But how do I justify this? From wikipedia, one of the things to see is that each variable should also have a normal distribution (e.g. the whole set of x1 is normally distributed, similar for x2 and x3 respectively). I have tried: normplot(All-X'es) ... and for x1 and x2 and x3, it seemed to fit across the diagonal line generated. Does this mean that my data is multivariate normally distributed? Hope I am not confusing anyone too much :) Thanks in advance Alf
From: TideMan on 20 Apr 2010 06:02
On Apr 20, 7:26 pm, "Alfian Abdul Halin" <jawatrox...(a)gmail.com> wrote: > Hi all, > > I was wondering if anyone knows how to determine that a particular set > of multivariate data is normally distributed? (Hope I'm using the multivariate term correctly, since I have just learnt it...) > > For example... I have a collection of feature vectors, where X = (x1,x2,x3)... > Overall, I am making N-observations... so X1, X2, ..., XN feature vectors are generated. > My assumption is that (based on some domain knowledge), a multivariate normal distribution will result. But how do I justify this? > > From wikipedia, one of the things to see is that each variable should also have a normal distribution (e.g. the whole set of x1 is normally distributed, similar for x2 and x3 respectively). I have tried: > > normplot(All-X'es) ... and for x1 and x2 and x3, it seemed to fit across the diagonal line > generated. Does this mean that my data is multivariate normally distributed? > > Hope I am not confusing anyone too much :) > > Thanks in advance > > Alf As a first check of Gaussianity, the 3rd and 4th statistical moments (skewness and kurtosis) need to be zero. If they are non-zero, then it is not a Gaussian distribution. |