Thanks to Kit Baum's adding the arimaauto command to the SSC, the class is publicly available after its installation.
More in this post. You will need to install hegy (not hegy4!) and kpss commands.
The Mata code:
https://github.com/econcz/stata-arim...arimaauto.mata
The certification file:
https://github.com/econcz/stata-arim...rtification.do
Comments:
ARIMAAuto (stored in the larimaauto.mlib file) stores several types of variables in an associative array (of which it is itself a child) declared with:
Contents of variables are stored and read with:
where "key" is the name of the variable.
The variables are varlist - [varlist] (string scalar), ifin - concatenated [if] [in] (string scalar), iw - [weight] (string scalar), o_* - options (string scalar), f_* - flags (real scalar), level - level (real scalar), mode and ic - information criteria (string scalar), and finally L, T, and MS - real matrices of limits, tests, and models.
NB The variables should be declared in the above given order.
L is a (1 x 8) matrix/vector of which arimaauto's r(limits) is a transposed copy, i.e. r(limits) = AA.get("L")'
T is a (. x 6) matrix of which arimaauto's r(tests) is a copy, i.e. r(tests) = AA.get("T")
MS is a (. x 12) matrix of which arimaauto's r(models) is a reduced copy, i.e. r(tests) = AA.get("MS")[,(1,3,4,6,8,10,11,12)]. The left out columns are #d, #D, #s, and constant (in the given order).
NB The example of their use can be found in the arimaauto.ado file.
PS ARIMAAuto includes a virtual function get_cv_seas() returning a real matrix of critical values (test x 0.01,0.05,0.10) for the hegy test (second row is retrieved), which can be replaced in an eventual child class when a different test is required (the second row rule must be observed, for example, coding AA.put("T", J(1,6,.)) in the beginning) together with the other variables). ARIMAAuto considers unit root to be present if statistic > critical value, hence signs may need to be adjusted for unit root tests other than hegy, [TS] dfgls, and kpss.
PSS All estimations in ARIMAAuto are performed under version 13.
More in this post. You will need to install hegy (not hegy4!) and kpss commands.
The Mata code:
https://github.com/econcz/stata-arim...arimaauto.mata
The certification file:
https://github.com/econcz/stata-arim...rtification.do
Comments:
ARIMAAuto (stored in the larimaauto.mlib file) stores several types of variables in an associative array (of which it is itself a child) declared with:
Code:
. mata: AA = ARIMAAuto()
Code:
. mata: AA.put("key", value)
. mata: AA.get("key")
The variables are varlist - [varlist] (string scalar), ifin - concatenated [if] [in] (string scalar), iw - [weight] (string scalar), o_* - options (string scalar), f_* - flags (real scalar), level - level (real scalar), mode and ic - information criteria (string scalar), and finally L, T, and MS - real matrices of limits, tests, and models.
NB The variables should be declared in the above given order.
L is a (1 x 8) matrix/vector of which arimaauto's r(limits) is a transposed copy, i.e. r(limits) = AA.get("L")'
T is a (. x 6) matrix of which arimaauto's r(tests) is a copy, i.e. r(tests) = AA.get("T")
MS is a (. x 12) matrix of which arimaauto's r(models) is a reduced copy, i.e. r(tests) = AA.get("MS")[,(1,3,4,6,8,10,11,12)]. The left out columns are #d, #D, #s, and constant (in the given order).
NB The example of their use can be found in the arimaauto.ado file.
PS ARIMAAuto includes a virtual function get_cv_seas() returning a real matrix of critical values (test x 0.01,0.05,0.10) for the hegy test (second row is retrieved), which can be replaced in an eventual child class when a different test is required (the second row rule must be observed, for example, coding AA.put("T", J(1,6,.)) in the beginning) together with the other variables). ARIMAAuto considers unit root to be present if statistic > critical value, hence signs may need to be adjusted for unit root tests other than hegy, [TS] dfgls, and kpss.
PSS All estimations in ARIMAAuto are performed under version 13.
Comment