Hi Isi, Do I understand correctly that you are proposing to implement yourself the points 1 and 3 ? Let me know if you do it. About your point 2, I will be happy to include the relevant piece of code if somebody provides it. Rene Brun Isard_Dunietz wrote: > > Dear Rene, > Probably the following features exist in root. I just > have not yet figured out how to get at them. > > 1) It is often useful to plot the data as a > "normal quantile plot" or sometimes called a "normal plot". > > Description of "normal plot" from the book by Tamhane and Dunlop, > Statistics and Data Analysis, p. 123, is given here: > > Suppose the data follow an N( mu, sigma^2 ) distribution, then the > percentiles of that normal distribution should plot linearly against the > sample percentiles, except for sampling variations. For the sample > percentiles one uses normally the ordered data values themselves, the i'th > ordered data value being the > 100( i/(n+1)) th sample percentile, where n is the sample size. The > corresponding standard normal percentiles are called normal scores. The > plot of these normal scores against the ordered data values is the normal > plot. > Is there some tool in root to plot out this normal plot? > > 2) It also would be useful to calculate > the inverse > cummulative density function (c.d.f.), i.e. > the inverse fnct of > cdf( z ) = ( 1 + TMath::Erf( z / TMath::Sqrt(2) ) ) > Is such a fnct easily obtainable inside root? > > 3) Also, is there a way to plot out data as a "box plot" or sometimes > called a "box and whiskers plot", which plots out the following > five important numbers > [either bin by bin, or for each sample separately]: > { x_min, Q_1, Q_2, Q_3, x_max }, where x_min, x_max are the > minimum/maximum > in the current sample and Q_k are the k'th quartile > (25percentiles). Please, > see, for instance, p. 121, Tamhane and Dunlop's book. > > Thank you for your guidance. Cheers, Isi
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:38 MET