[Gretl-users] Problem with matrix command

Hello listers ! greetings from Paris ! I am currently working now on a script that i will soon submit to the community. However, I think i have spotted a bug with possibly serious consequences. I am under WinXP running on 1.7.8 (not snapshot) The script rely heavily on matrix commands and SVDs. I was puzzled to see that the results provided by GRETL on certified data (ie, the data from papers, so that i can be sure to have the good results) where completely at odds from those of the paper ! Verifying the commands, I checked for any mistake left in the code. All the formulae where ok, and conform to those exposed in the paper. Then, as the data matrix is badly conditionned, (CI > 10^17...) i wondered if there could be a numerical problem around (svd algorithm etc). Finally is spotted the source of the error: the matrix per element exponentiation seems to be falty. Try the small script : matrix T = {1,0,0,0;0,0.5,0,0;0,0,1.5,0;0,0,0,0.125} T matrix S = inv(sqrt(T)) S matrix V = T.^-0.5 V matrix ZZ = S/V ZZ normally matrix, S whose elements are equal to : 1/sqrt(Tij) should be the same as in matrix V. However, this is not the case and this was the cause of my concern. Now i use matrix S command and the results are in line with the papers i am reproducing. I think something may need to be fixed here. thanx and cheers ! Franck