Dear listers, hi from Paris ! I have a problem with matrix code. the following is part of my code : # standard errors of elasticities matrix EP={EP11,EP12,EP13;EP21,EP22,EP23;EP31,EP32,EP33} matrix EP matrix ER1 matrix ER2 matrix ER3 matrix ER = {ER1;ER2;ER3} matrix ER The EP matrix is built without any problem but not the ER matrix : ===========cut off ouput for RP=================== =============ER matrix faulty execution=========== ? matrix ER1 ER1 0.0038225 ? matrix ER2 ER2 0.011485 ? matrix ER3 ER3 0.00063132 ? matrix ER = {ER1;ER2;ER3} Error adding matrix 'ER' Unknown variable name in command Error executing script: halting > matrix ER = {ER1;ER2;ER3} ======================================================= I dont see what is wrong : the EP matrix is OK but not the second one, ER. Note that both matrices are built from existing scalars (actually results of prio matrix operations). As can be seen, matrices ERi are defined but the matrix command rejects it as a valid variable name. Thanx for your help Franck _______________________________________________ Gretl-users mailing list Gretl-users(a)ricardo.ecn.wfu.edu http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users

Hi Sven thanx for your quick response. but err... It does not work too... In fact what is puzzling is that i changed the code because the ER1, ER2, ER3 are actually scalars. So I changed the code to declare them as scalars, then as can be seen i show them again (to check they are correctly defined and accessible), so that i build the matrix ER = {ER1;ER2;ER3} from them but gretel says they are undefined. Now I change the code so that they are computed as matrices, so that i can use your advice by concatenating them later... drums sounding... ooouch... does not work : ======================================================= # elasticities standard errors matrices ? matrix EP={EP11,EP12,EP13;EP21,EP22,EP23;EP31,EP32,EP33} Replaced matrix 'EP' ? matrix EP EP 0.0095979 0.0065156 0.0068756 0.028839 0.019577 0.020659 0.0015852 0.0010761 0.0011356 ? matrix ER = (ER1 ~ ER2 ~ ER3)' Error adding matrix 'ER' Error executing script: halting > matrix ER = (ER1 ~ ER2 ~ ER3)' =========================================== The problem is still there : matrix EP is correctly built but not ER. The elements of both are still accessible elsewhere... Well, help very welcome ! cheers Franck

> What version of gretl are you using? > > I have gretl "official" 1.6.0 in Linux and I obtain the same error, but I > installed the last gretl Windows snapshot in other computer and the > command from Sven works ok. Well I have version 1.6.0, build 9-11-2006, for windows. cheers

matrix Z[1,]=transp(logement.$coeff[2

matrix Z[1,]=transp(logement.$coeff[2:5])

? matrix Eq1=logement.$coeff Generated matrix Eq1 ? matrix Eq1 Invalid declaration

logement <- ols WLOG const l_PLOG l_PTRANS l_PCOMP lnREALEXPCS transport <- ols WTRANS const l_PLOG l_PTRANS l_PCOMP lnREALEXPCS composite <- ols WCOMP const l_PLOG l_PTRANS l_PCOMP lnREALEXPCS # then i want to extract the coefficients 2 to 5 of each equation: matrix Z[1,]=transp(logement.$coeff[2:5])

Thanks for your patience! It's very useful to get these bug reports. But I think the issue is sorted out in CVS, and I'm putting a new Windows snapshot in place now.

You should now be OK with what you were trying originally, such as matrix B = transp(mymodel.$coeff[2:5]) You should also be able to make a selection from the VCV matrix, as in matrix S = mymodel.$vcv[2:5,2:5] Allin.

On Tue, January 30, 2007 12:56, Sven Schreiber wrote: > (btw, a vertical concatenation operator wouldn't hurt ;-) I was thinking the same thing a few weeks ago. The one thing that stopped me from implementing this is that the "|" character, almost unversally used to indicate vertical concatenation in matrix languages, is already taken by the "or" operation. It it feasible, in principle, to distinguish between the two uses: theoretically, given the expression "a | b" we first check that a and b are matrices with the same number of columns; if not, "|" should be interpreted as "or". I fear this could end up being a bit confusing, though.

Allin Cottrell schrieb:

> I've added vertical concatenation in CVS, for the moment using '`' > (backtick) as the operator symbol. If anyone has a better idea, I'll be > happy to listen. Backslash or '#' would be other possibilities. Does it have to be one char? Otherwise what about ~' which I find quite intuitive: vertical concat. == "transposed" horiz. concat.

> Sorry, everyone, I don't mean to spoil the fun, but ~' doesn't exactly excite me.

Hence, my counter-proposal (which, I admit, does have drawbacks, but is still preferable IMHO): use "|" for vertical concatenation and switch to "&&" and "||" for logical operators. It will break existing syntax, but it'd be better in the long run.

I have a problem with matrix code. the following is part of my code : # standard errors of elasticities matrix EP={EP11,EP12,EP13;EP21,EP22,EP23;EP31,EP32,EP33} matrix EP matrix ER1 matrix ER2 matrix ER3 matrix ER = {ER1;ER2;ER3} matrix ER The EP matrix is built without any problem but not the ER matrix...