Mixed Data Sampling in Stata (MIDAS) - more info needed

[Marian Dobranschi]

Dear Stata Specialists,

I had raised the same question in a previous post, but was somehow ignored.
I'm kindly asking for information regarding the MIDAS regression development in Stata. To be more clear and to avoid any confusion, MIDAS regression is NOT the same as Midas used in Stata for meta-analysis of binary diagnostic test accuracy. Mixed Data Sampling or MIDAS regression is a rather new topic in statistics software where different frequency data sampling are used in the same regression.
In other words, variables that are included into a regression are combined frequency: daily, weekly, monthly and yearly. It is a rule of thumb that in the traditional methods the estimation of dependent and independent variables should use the same frequency: daily to daily or yearly to yearly and no combinations of higher or lower frequency are allowed. This rule becomes even more restrictive when panel data is used (to achieve strongly balanced panels). In comparison with Stata, there are some developments in Matlab for MIDAS regression, but for a beginner like me Matlab is way above my IT and statistics knowledge.
Is there someone that used MIDAS regression in Stata, where different data frequency has been introduced in the same equation? Moreover, is there a possibility to apply MIDAS regression to the panel data analysis? The main reason for this whole argument is due to the lack of same frequency data, where some data is reported daily and other data is reported quarterly or yearly, thus the researchers are forced to either aggregate and lose the accuracy or use MIDAS regression.
Thank you very much for your time and attention.

[Carlo Lazzaro]

Marian:
as far as I know, -search MIDAS, all- gives back one entry only, exactly the one you're not looking for.

[Nick Cox]

I agree with Carlo. It appears that no-one here knows anything about an implementation of this method in Stata.

FAQ Advice includes comments on the otherwise theological issue of why questions go unanswered:

17. Questions that do not get answered

Questions can get no replies for many different reasons. Here, however, are some common ones.

• No one knows of any such Stata program. You may need to write your own code or use some other software.
• Your question really should be answered by using the help, the manual, or by typing search in an up-to-date Stata.
• We do not have the knowledge of your project needed to work out the best thing to do in your circumstances, and, in any case, it is really your call.
• Whether what you are doing is “correct” is very difficult to discuss helpfully.
• You did not provide enough information. For example, postings of the form “I tried using foobar, but it did not work” are usually impossible to answer, except by asking for more information.
• Your question is too unclear or too complicated to understand. For example, questions on very complicated data-management tasks or large chunks of code that are not working may ask too much.
• A model may not converge or fit well because it doesn't suit the data, or if you prefer the data don't suit the model. It can be very hard to advise on such cases, especially if presented generally.

[Marian Dobranschi]

Many thanks for your replies. My opinion is that this post should be open to discussion, maybe from all members of this Forum there is someone that tried this method and has some clues.
With all due respect, I think that references to FAQ Advice are not constructive neither helpful in solving this issue. I also think I was crystal clear in describing the problem and raising the relevant questions.

[Rich Goldstein]

it would have helped if you had provided some guidance/help to readers; I had never heard of this, but a quick google search found the following cite: Eric Ghysels, Sinko, A., Valkanov, R. (2007) MIDAS Regressions: Further Results and New Directions. Econometric Reviews, 26 (1), 53–90; is this what you are referring to?

[Nick Cox]

Marian: I think you're misunderstanding. If you re-read the FAQ, you'd see the quoted section in context as generic advice.

1. Your post remains entirely open to discussion. If someone comes along and tells us about an implementation of the technique in Stata, everyone will be happy.

2. Your question was to me clear. If you had added precise literature or internet references. that would have done no harm. But the explanation

"Mixed Data Sampling or MIDAS regression is a rather new topic in statistics software where different frequency data sampling are used in the same regression."

should be entirely sufficient to evoke recognition if such a program exists in Stata. Carlo used search in Stata and found nothing, as reported.

3. You appeared to be raising a secondary question in commenting that your post was "somehow ignored". Similar comments have often been made by puzzled list members, which is what that why advice exists in the FAQ. The misunderstanding sometimes arises because people think that every question will get an answer, which is promised nowhere. As said, it's generic advice and most of it has indeed no bearing on your question, but the first point, on the evidence so far, is completely pertinent.

[Marian Dobranschi]

it would have helped if you had provided some guidance/help to readers; I had never heard of this, but a quick google search found the following cite: Eric Ghysels, Sinko, A., Valkanov, R. (2007) MIDAS Regressions: Further Results and New Directions. Econometric Reviews, 26 (1), 53–90; is this what you are referring to?

Exactly this is one of the many articles on MIDAS Regressions.
The first article written on this method by Eric Ghysels is this one:

• Ghysels, E., Santa-Clara, P., Valkanov, R., 2002b. The MIDAS touch: mixed data sampling regression. Discussion Paper UCLA and UNC available at: http://www.unc.edu/eghysels.

[jerome falken]

I'm also interested seeing this addition to Stata.

[Nicolas Maystre]

Dear Stata users,
Similarly to other user above, I would be very interested in seeing this tool implemented in Stata in the future.
Best,
Nicolas

[Anand Singh]

I am also working with Mixed Data Sampling and looking for this power full technique in Stata, else we need to go to MATLAB in future. Is there anyone who can help with STATA programing for this technique.

[Xueliang Liu]

Me too, I have a project in which mixed frequency data is used. It seems there are packages for mixed frequency data in MATLAB, R and Eviews. But really don't want to use these tool, I am more used to stata now. It has been nearly 10 years since the last time I use Matlab and Eviews.

[John Sgr]

Dear Nick,

Is it appropriate to use interpolation to solve this different data frequency problem? I also have a similar problem, in my data I have monthly observations for depression score (dependent variable) and weekly observations for income level (independent variable). Therefore, I aggregated the weekly observation on a monthly basis to obtain a balanced dataset. However, I am wondering if there is an alternative method to solve this frequency problem?

Also, what do you think about the "ascol" command? (link: https://fintechprofessor.com/ascol-s...y-and-monthly/)

Best,
John

[Mike Kraft]

Dear All
Is the mixed data sampling regression available now in Stata?
Best

[Jiahao Liu]

Dear All
Is the mixed data sampling regression available now in Stata?
Best

It is now avaliable. The "midasreg" command written by Prof. Qunyong Wang from Nankai University, China, is now avaliable. But the command is for private, and you have to pay for the files only in RMB. link: https://bbs.pinggu.org/thread-8545491-1-1.html