Possibility to create group of "depends" variables from equation

[farrel]

Dear Eviews,

I would suggest to create function, which would create a group of variables from equation on which this group depends. This is sort of a mix of functions "makeregs" and "@depends". What I mean?
Assume you have an equation:

equation1.ls INFLATION INFLATION(-1) INFLATION(-2) CRUDEOIL CRUDEOIL(-1) CRUDEOIL(-2) @during("2008m12") @trend

New function, let's call it "makedepends", would create a group of the following variables: INFLATION CRUDEOIL @during("2008m12") @trend
i.e. only those variables, on which equation depends on!

Unlike "makeregs" function, new function eliminates all possible lags. (Imagine rich ARDL model, which determines lags automatically, and using programming it's difficult to extract only variables which "makeregs" depends on)

Another way would be to use the function "@depends" in equation1 to extract variables on which this group depends on. For example, let's make group of regressors "mygroup", then make a string of depends using @depends and transform the string to a new group, as I show below:

equation1.makeregs mygroup
string mystring = mygroup.@depends
group group_{mystring}

But I find the function "@depends" a bit weird. Why?! because in case of equation1, the result would the following group:
group_inflation INFLATION INFLATION INFLATION CRUDEOIL CRUDEOIL CRUDEOIL @during("2008m12") @trend

I mean, why I need series INFLATION and CRUDEOIL 3xtimes in one group?!

[EViews Gareth]

Neat idea.

[farrel]

Neat idea.

But what about to implement?! :))

[EViews Gareth]

It will be in the next major release

[farrel]

It will be in the next major release

Gareth, thank you!

I have 2 more things in the same context. Gareth, could you please to have a look at this?

1. The FIRST

I have tried to extract series from the "makeregs" group and then using @depends function to obtain series, but something doesn't work. Look at the following code using "equation1" from the post above:

Code: Select all

{%equation}.makeregs gr_regs ' make regressors group %str_regs = gr_regs.@depends ' create string of dependent vars group gr_{%str_regs} ' create a group of dependent vars %depreg = gr_{%str_regs}.@seriesname(1) ' make a string of the first series in the group of dep.vars show {%depreg} 'show the first variable in the group of dep.vars

I expect this code to work, but it doesn't. Eviews says that there is an error in "%depreg" (row 4), the string cannot be compiled. But to me, the code seems ok and should work ...

2. The SECOND

About the sequence of variables in the output of ARDL model. Assume ARDL model:

Code: Select all

EQ1.ardl(fixed, deplags=2, reglags=1) LOG(HICP) LOG(WAGESA) LOG(FOOD) @DURING("2005m07")

The sequence of variables in the ouput is the following (and the same sequence using makeregs function):
LOG(HICP) LOG(HICP(-1)) LOG(HICP(-2)) LOG(WAGESA) LOG(WAGESA(-1)) LOG(FOOD) LOG(FOOD(-1)) @DURING("2005m07")

Everything is fine. The sequence of variables is the same in both equation and output.

But if I define EQ1 with fixed lags using @fl() function the sequence of variables changes. Here is equation:

Code: Select all

EQ1.ardl(fixed, deplags=2, reglags=1) LOG(HICP) @FL(LOG(WAGESA),0) LOG(FOOD) @DURING("2005m07")

and here output (or the group using makeregs):
LOG(HICP) LOG(HICP(-1)) LOG(HICP(-2)) LOG(FOOD) LOG(FOOD(-1)) LOG(WAGESA) @DURING("2005m07")

Note that using @fl() function, the variable LOG(WAGESA) goes after LOG(FOOD).
From programming point of view it's disappointing, so I cannot keep the sequence ov variables defining equation and defining the group of variables in this equation.
Is it possibel to fix it?

[EViews Gareth]

1) %str_regs has a space in it, so can't be used as a variable name.
2) Unfortunately not.

[farrel]

1) OK I see. I overcome it by using string vector and split string in parts.
2) I hope a new "makedepends" function will fix it ...

[farrel]

It will be in the next major release

Gareth, have you managed to introduce this function in Eviews12 we talked a year ago?

Andrejs

[farrel]

Is it still in a pipeline?

Andrejs

[EViews Gareth]

Actually, I take it back - it is in EViews 12.

Code: Select all

eqname.@depends

[farrel]

Great. thanks!
Even though the function doesn't reflect dummies [@during()] and trend (@trend). I would add them.