GitLab

Fixed unitary tests. @dynDates/sort and @dynDates/unique methods have to be called without (empty) parenthesis.

*Example* Let dd be a dynDate object:

d1 = dynDate('1966Q4');
d2 = dynDate('2013Q1');
dd = d1:d2;

then dd(2:5) is also dynDates object for the dates 1967Q1, 1967Q2, 1967Q3 and 1967Q5.

*Example* If a dynDates object is built as follows

d1 = dynDate('1966Q4');
d2 = dynDate('2013Q1');
dd = d1:d2;

then dd(2) is a dynDate object (should be equal to dynDate('1967Q1')).

Call dynDate constructor with two arguments: teh first one is a 2*1 vector of integers (for time), the second one is a scalar integer (for freq).

If ts is a dynSeries object, then the following syntax is valid:

t0 = dynDate('1950Q1') ;
t1 = dynDate('2000Q3') ;

range = t0:t1 ;
ds = ts(range);

Note that the dates in range do not need to be contiguous or sorted in increasing order.

option using second order perturbation correction,
solve_stochastic_perfect_foresight_model.m moves to matlab/ep directory

Handle all events that lead to a penalty through an exception.
Next steps:
- make all exceptions derive from a common class containing info code and
  penalty value
- catch these exceptions at the top-level, and construct the penalty-adjusted
  likelihood

Does not seem very robust (fails on fs2000), need to investigate why.

In the calling chain, only pass the basename of the MOD file.

We now have the `calib_smoother' command.

Speeding up of forecasting, cosmetic change

Deleted redundant global statement and initialized arrays used