Thanks Johannes,
it seems that deterministic models with several leads and lags and various combinations of initval, histval and endval have never been properly tested. In addition, since the addition of histval, the meaning of endval has become ambiguous. The reason to exclude histval and endval in the same *.mod file is that initval is free to declare the terminal condition but I agree that it is extremely confusing
Best
Michel
Johannes Pfeifer writes:
Dear Michel,
your commit did not fully solve the issue. See the two attached mod-files. Please ignore the fact that the results will return NaN due to histval not being fully initialized.
For the ramst.mod due to histval being present, there are more than two lags added to the vectors, again leading to a crash in the reshape.
For the retp1 the length of endo_val and exo_val is not even conformable.
I have an additional question: why can one not use endval and histval simultaneously (enforced in make_y_ )? I thought the first one specifies the lags and the other one leads? This restriction makes it hard to enter a permanent shock while providing values for lags. Moreover, we should document this restriction in the manual.
Best,
Johannes
--
Johannes Pfeifer
Haußerstr. 29
72076 Tübingen
Tel.: +49-(0)7071-6396184
Mobil.: +49-(0)170-6936820
Germany