### Dynamic Model Averaging

Posted:

**Tue Sep 06, 2016 2:51 pm**This thread is about the dma add-in that performs Dynamic Model Averaging (Koop and Korobilis 2012).

Page **1** of **2**

Posted: **Tue Sep 06, 2016 2:51 pm**

This thread is about the dma add-in that performs Dynamic Model Averaging (Koop and Korobilis 2012).

Posted: **Mon Sep 25, 2017 9:23 am**

Hi,

may I ask: in the example file is a command line of the code as such

dma(phlag="2 0", sample="1960q1 2008q2") gdpdefl 4 5 t_code @ unemp hstart employ tbill spread djia money infexp comprice vendor

I am confused about what the 4 5 digits mean. I tried estimating the model using the combo box but I keep getting an error msge: "3 is not a valid index for vector-series-coefficient tcode". I am estimating model with 2 exogenous variables

Thank you

Martin

may I ask: in the example file is a command line of the code as such

dma(phlag="2 0", sample="1960q1 2008q2") gdpdefl 4 5 t_code @ unemp hstart employ tbill spread djia money infexp comprice vendor

I am confused about what the 4 5 digits mean. I tried estimating the model using the combo box but I keep getting an error msge: "3 is not a valid index for vector-series-coefficient tcode". I am estimating model with 2 exogenous variables

Thank you

Martin

Posted: **Mon Sep 25, 2017 3:41 pm**

I am confused about what the 4 5 digits mean.

4 is forecast horizon. 5 is transformation code (log difference) for dependent variable.

I tried estimating the model using the combo box but I keep getting an error msge: "3 is not a valid index for vector-series-coefficient tcode". I am estimating model with 2 exogenous variables

the following code with 2 exogenous variables is working.

Code: Select all

`dma(phlag="2 0", sample="1960q1 2008q2") gdpdefl 4 5 t_code @ unemp hstart `

read the instruction pdf file.

Posted: **Sat Sep 30, 2017 9:27 am**

Dear dakila,

I made a mistake contructing the vector t_code for my own exercise. I did not know that the dependent variable is not to be included in the vector of transformations.

Thank you for this awesome add-in !

I made a mistake contructing the vector t_code for my own exercise. I did not know that the dependent variable is not to be included in the vector of transformations.

Thank you for this awesome add-in !

Posted: **Tue Apr 10, 2018 11:14 pm**

Hello, I have Eviews 9.5. I downloaded the dma add-ins and installed it but canâ€™t find any supporting materials with the instructions of how to use it. particularly, what is the vector transformation code? many thanks

Posted: **Wed Apr 11, 2018 12:12 am**

The instruction pdf file is located in C:\Users\...\Documents\EViews Addins\DMA\dma.pdf

code explanation

1 Level

2 First Difference

3 Second Difference

4 Log-Level

5 Log-First-Difference

6 Log-Second-Difference

code explanation

1 Level

2 First Difference

3 Second Difference

4 Log-Level

5 Log-First-Difference

6 Log-Second-Difference

Posted: **Wed Apr 11, 2018 2:04 am**

Hi dakila, I am sorry my ignorance, but that file should be located in my C: folder after I installed the add-in?

I can't find that file name in my machine.

Or is it located in this website?

Thanks,

Carolina

I can't find that file name in my machine.

Or is it located in this website?

Thanks,

Carolina

Posted: **Wed Apr 11, 2018 4:27 am**

Posted: **Thu May 24, 2018 6:16 am**

Hi dakila,

Can you explain what is the difference of the 3 outputs we get from your addin?

My dependant variable is y and after running the DMA program I get y_t_, y_t_best01, y_t_dma01

Thanks

Can you explain what is the difference of the 3 outputs we get from your addin?

My dependant variable is y and after running the DMA program I get y_t_, y_t_best01, y_t_dma01

Thanks

Posted: **Sat May 26, 2018 3:46 pm**

y_t_ is the transformed dependent variable.

y_t_best is dynamic model selection forecast (best model forecast).

y_t_dma is dynamic model averaging forecast.

y_t_best is dynamic model selection forecast (best model forecast).

y_t_dma is dynamic model averaging forecast.

Posted: **Tue Jul 03, 2018 2:39 am**

hi dakila, why my resulted transformed variable is different from the dependent variable even if the transformation code used is 1 (level) and a forecasts horizon of 6 (months)?

I would like to transform back the best forecasts series but I can't understand how y_t was transformed initially and why.

many thanks

I would like to transform back the best forecasts series but I can't understand how y_t was transformed initially and why.

many thanks

Posted: **Tue Jul 24, 2018 7:44 am**

That is bug in the add-in. I will fix it soon.

Posted: **Tue Jul 24, 2018 7:53 am**

Hi Dakila,

I have seen that the resulted Y_T, Y_OUT_BEST and Y_OUT_DMA are exactly TWO TIMES the value of the original variable. When I do Out of Sample forecasts should I assume that Y_T_BEST01 and Y_T_DMA01 are also two times the value of the correct forecast?

Additionally, is there a way of getting the charts for the changing coefficients across time? So I can see how the latest forecasts are calculated.

Thanks,

Carolina

I have seen that the resulted Y_T, Y_OUT_BEST and Y_OUT_DMA are exactly TWO TIMES the value of the original variable. When I do Out of Sample forecasts should I assume that Y_T_BEST01 and Y_T_DMA01 are also two times the value of the correct forecast?

Additionally, is there a way of getting the charts for the changing coefficients across time? So I can see how the latest forecasts are calculated.

Thanks,

Carolina

Posted: **Tue Jul 24, 2018 8:38 am**

sorry I don't understand your question. What is the original variable. Is it left hand-side variable(y_t)? take an example.

in the future I will try to include the changing coefficients.

in the future I will try to include the changing coefficients.

Posted: **Tue Jul 24, 2018 8:51 am**

My left hand side variable name is FTSE_REL, the resulted output gives a Y_T_ series that is exactly twice the value of FTSE_REL, even though no transformation is instructed in the code.

Moreover, Y_T_BEST and Y_T_DMA are the same as Y_T_ (twice FTSE_REL) for the whole sample period until the end of the sample period (2018m06), after that, they provide two different set of forecasts (6M forecast horizon in my case).

Additionally, I get Y_T_BEST01 and Y_T_DMA01 that are different that Y_T_ for the entire sample time period but do not provide forecasts after 2018m06.

Would you mind explaining to what corresponds each of these resulted series?

Moreover, Y_T_BEST and Y_T_DMA are the same as Y_T_ (twice FTSE_REL) for the whole sample period until the end of the sample period (2018m06), after that, they provide two different set of forecasts (6M forecast horizon in my case).

Additionally, I get Y_T_BEST01 and Y_T_DMA01 that are different that Y_T_ for the entire sample time period but do not provide forecasts after 2018m06.

Would you mind explaining to what corresponds each of these resulted series?