EViews.com

Dear Gareth,

I am reaching out to seek assistance with a technical issue I encountered while using EViews for simulations in an ARDL model. After successfully estimating the model and performing an initial simulation without any issues, I am experiencing a problem during a specific second simulation scenario.

Here is a summary of the procedure I followed:

Estimation of an ARDL model with variables mortgage_rate, ger_3m_yield, log(qe), mp_surprises, and france_inflation for the period from July 2009 to 2023.
Conducting baseline simulations (deterministic and stochastic).
Implementation of two simulation scenarios: the first proceeds without issue, but during the execution of the second scenario, EViews displays the following error: "Unable to compute due to missing data in 'MORTGAGE_RATE = @COEF(1) *...'".
It is important to note that this error occurs despite:

The ARDL model successfully passing all statistical and stability tests.
A manual simulation following the second scenario, implemented with an alternate code, working correctly.
Attached, you will find the EViews script I used, including the ARDL model estimation and the two simulation scenarios. I am particularly puzzled as to why the automated simulation fails while the manual simulation succeeds, especially in the absence of data issues or model specification problems.

I would be extremely grateful if you could review this issue and provide any guidance or potential solutions. I am available for any further information or clarification needed.

Thank you in advance for your time and valuable assistance.

Best regards,

Bob

qe_sc2 needs values for the pre-sample periods (due to lags in the model).

Following your response regarding the need for pre-sample values for qe_sc2 due to lags in the ARDL model, I would like to clarify. My ARDL model (3,0,3,1,2) is estimated from July 2009 to December 2023, and the simulations are for January 2020, suggesting sufficient pre-sample data should be available.

To assist in clarifying and verifying this, I have included the database and program as attachments. Could there be another aspect I might have overlooked that could explain the issue encountered with the second simulation scenario?

Thank you for your assistance.

If you open the qe_sc2 series, over which dates does it have data?

Thank you, Gareth, you were right, it was necessary to use a 'smpl @all' before creating qe_sc2 and then return to the desired smpl afterwards! I mistakenly thought that the smpl declared before the model creation was okay. Thanks again !!!

Still related to the program, for example, for sc1, I indeed observe the 'deviation' between sc1 and the baseline, but not the 'percentage deviation, is it normal ?