[Daniel Feenberg]

1) Support for column oriented Parquet data files as an alternative to .dta format for Stata binary files. When only a fraction of variables are needed, this is very efficient. See https://www.databricks.com/glossary/what-is-parquet for a discussion of why/when column orientation is appropriate.

2) A way to determine the maximum memory used since job start. This is useful because many Stata procedures use temporary memory that is given up between procedures. So the -memory- command isn't helpful in learning what amount of memory is needed.

3) -use- and -net from- should follow http redirects. See https://curl.se/docs/comparison-table.html to see how common this feature is among libraries supporting http access.

4) Error messages from within -for- loops should contain information about which code line triggered the message.

5) -save- should support -if- and variable lists. -save- should allow http targets.

[Bruce Weaver]

The default (or factory setting) scheme in v18 is stcolor. I surprised me that there is (presently) no stmono. I would like to see that added (perhaps even in an update to v18).

• https://www.stata.com/manuals/g-4schemesintro.pdf

[FernandoRios]

Perhaps they can consider adding some tools so users can write functions. This is rather simple in mata, but cannot figure out a way to do it with Stata itself.
for instance, it would be great to be able to write something like
gen y = fun1(x1) + fun2(x2)
where fun1 and fun2 are defined by me.

Thank you
F

[FernandoRios]

From the Graphics perspective.
Perhaps expand on "twoway" capabilities so it is easier to program extensions for it. That way it would allow for:

twoway mygraph x1 x2 || scatter x1 x2

F

[FernandoRios]

One more!
Some time ago, Stata introduced Numerics as a web-based implementation that would suit particular needs some time ago. Unfortunately, there is no example of how it would work. The example cited in the website "ADEPT" is no longer active, so it's hard to figure out how it would work.
Thus, It would be good if they could provide small examples of how Numerics works.
F

[KitBaum]

I second Bruce Weaver's request for stmono. There should be a monochrome scheme with the same defaults for background, axis labels, etc. for those cases where you need to produce a monochrome graph, without all of the old-style features of s2mono. There are plenty of community-contributed schemes that are quite similar to stcolor, but there should be an official Stata monochrome scheme. I'm surprised the Stata Journal production staff haven't made this request.

[Christopher Bratt]

Adding to #18: In Stata, programs have to be coded as ado-files, they cannot be integrated in a do-file as part of a specific workflow, right? (I can’t find any reference to the option to write a program within a do-file, with the do-file completing many other tasks.)

If that’s correct, i. e., all user-defined Stata programs have to be written as ado-files, what about Stata 19 allowing for the writing of programs within a do-file, with the option for the user to call that program within/later in the do-file? The program would be in memory only, activated only by the computer reading the particular do-file.

That would resemble the use of functions in R, which can be declared within an R-session to make long code more compact and easier to debug. It would also promote literate programming (my main reason for using R).

[Leonardo Guizzetti]

Adding to #18: In Stata, programs have to be coded as ado-files, they cannot be integrated in a do-file as part of a specific workflow, right? (I can’t find any reference to the option to write a program within a do-file, with the do-file completing many other tasks.)

If that’s correct, i. e., all user-defined Stata programs have to be written as ado-files, what about Stata 19 allowing for the writing of programs within a do-file, with the option for the user to call that program within/later in the do-file? The program would be in memory only, activated only by the computer reading the particular do-file.

That would resemble the use of functions in R, which can be declared within an R-session to make long code more compact and easier to debug. It would also promote literate programming (my main reason for using R).

You are under a misconception. -ado- files are not really more special than -do- files, they just give you the ability to elevate programs into commands. Both types of files can contain as many mata functions and Stata programs as you wish. There are also some details about scope, but that's perhaps a diversion from your main question. Therefore, coding reusable programs are absolutely possible, in either type of file. In this respect, Stata is "literate."

Something to note if you dive into this a bit more is that Stata tends to encourage reusability by using -ado-files. Code them once, place them in your ado-path, and you can access them without the need to explicitly import that code in a particular do-file. Maybe you don't want that, and instead you have written fit-for-purpose programs that are to be reused in one project, but are not so reusable elsewhere. That's fine too. You can still go down the -ado-path, or if you wish, it is sometimes easier to push programs into a dedicated do-file and then import the file as needed. (Or you can choose to write the style of spaghetti and pile all of your code into a monolithic do-file...but then again, that's not very "literate" in my humble view.)

[Christopher Bratt]

Thanks, Leonardo.

[daniel klein]

More on the background story of ado-files in Newton (2005, sec. 5)

Newton, J. H. 2005. A conversation with William Gould. The Stata Journal, 5(1), 19–31.

[JanDitzen]

1) Support for column oriented Parquet data files as an alternative to .dta format for Stata binary files. When only a fraction of variables are needed, this is very efficient. See https://www.databricks.com/glossary/what-is-parquet for a discussion of why/when column orientation is appropriate.

2) A way to determine the maximum memory used since job start. This is useful because many Stata procedures use temporary memory that is given up between procedures. So the -memory- command isn't helpful in learning what amount of memory is needed.

3) -use- and -net from- should follow http redirects. See https://curl.se/docs/comparison-table.html to see how common this feature is among libraries supporting http access.

4) Error messages from within -for- loops should contain information about which code line triggered the message.

5) -save- should support -if- and variable lists. -save- should allow http targets.

I strongly agree with the points. I would add on 3) the command copy - this would be extremely helpful. It would also be great if Stata would have it's own json read/write command. Regarding 5) are you thinking of writing to http targets? I believe then one would need ftp or similar?

[Bruce Weaver]

Please note that no disrespect of -corrci- (sj) or any other user-contributed packages is implied in this post!

I think it is high time that the following commands include options to report confidence intervals [1]:

• correlate & pwcorr
• spearman & ktau
• tetrachoric

Furthermore, it appears there is still no command to compute the biserial (NOT point-biserial) correlation. That would be a nice addition.


[1] FWIW, SPSS 29 has options to compute CIs for Pearson's r and Spearman's rho. See also:

• https://www.ibm.com/docs/en/spss-sta...dence-interval

[Fahad Mirza]

Can you elaborate please?

By that I mean allow Stata to read pixel data from images in PNG, JPEG, TIFF etc.

Gradient colors would be interesting as it allows for transition of state. Imagine a sankey plot with gradient tones.

[Fahad Mirza]

Please integrate GitHub with Stata so that users can set up and make use of repositories with ease. If this makes sense.

[Giorgio Di Stefano]

TVP, TVP- VARs and PANEL VARs, Panel TVP- VAR models.
It is not accepted those models to be solid and state of the art in macroeconomic research for two decades now and not be part of a serious statistics package. Plus in my direct inquire to stata corporation I I ve have given the reply that those models were/ are in their wish list, since when stata 14 come out! . It will be sufficient to imitate the BEAR toolbox from the ECB but be able to handle gaps