[Christopher Bratt]

Being able to set

 #delimit ;

permanently would help me a lot. But such an option would pose a problem for Stata code that others may use, so I assume StataCorp wouldn't allow a permanent change.

So, as an alternative:

Being able to set

#delimit ;

in interactive use. (I run Stata from R and RStudio, so Stata seems to regard my code as interactive).

[William Lisowski]

Some of us have to work in multiple environments - such as on our desktop, routinely, and in a secured enclave when working with sensitive data. Duplicating the Stata configuration of our primary environment in other environments is far from straightforward.

It would be useful if saved preference sets could the be exported and imported. It would be useful if there were a "smart" command that would produce a do-file with all the commands necessary to reproduce the changes (relative to the installation default) found in the output of creturn list. By a "smart" command I mean that I do not have Stata/MP so when I'm working on a Stata/MP system, I don't need c(processors) reduced to 1 to match my desktop system. :-) 1 is the default on non-MP systems, so the do-file should leave the default provided on the target system unchanged.

I know this can't be implemented perfectly, but the do-file I envision would be a starting point, something that I could review and edit as needed.

[Mario Ferri]

Please elaborate what you mean, giving some specific examples, for those of us who don't know R or Matlab intimately. As far as I understand an R "package" and a Matlab "toolbox" have analogues in Stata -- they are the programme or programmes (ado-files and help-files) that form a package of files. (Examples are the packages that one can download, for free, from SSC or the Stata Journal website.) R packages, e.g. those sitting on Cran, are typically free to download ... but many have to be user-downloaded (as in Stata). My googling re Matlab toolboxes suggests that one has to pay for some of them: see e.g. https://www.mathworks.com/help/thing...ox-access.html. So, another question is: what precisely do you mean by "custom"?

The one example you refer to is 3D graphics. I type -search 3D- in Stata and get a number of packages showing up, including contour plots and some 3D 'wire' diagrams. What functionality are you missing in particular? The ability to rotate 3D wire diagrams to show different perspectives, perhaps using a mouse? Or something else?

I don't want to miss out on any good stuff 'out there'; I am genuinely interested in learning about different ways of doing things. It's this that motivates my questions.

Current research in macroeconomics is done with custom codes and toolboxes. You tell the program what you want to do and how to implement it. You don't expect to be already written in some ado program or other.
There are toolboxes that users have to pay for but most of them are custom codes and toolboxes developed by researchers and are free to use. Usual platforms are R and Matlab. R is free. of charge. That is what I am referring. Integration with both those programs, and possibilities to run those free codes and toolboxes in Stata will sky rocket the program.
As for graphics, yes there are graphs but in the market there are other softwares competitive that are much better than Stata in graphs like R.

[wbuchanan]

Mario Ferri
In all honesty, there is much more than software needed to generate 3D representations of data. I suspect you’re not referring to a literal case, but are more interested in “3D” visualizations that have been projected down into two dimensional space (e.g., what you would see on the screen resulting from any commands). Aside from some of the suggestions that Stephen made above, Stata has already been able to visualize data in multi-dimensional space but does so by mapping additional elements to aesthetic attributes or features in the visualization. Contour plots are a perfect example of how three dimensions of data are mapped natively into two dimensions by mapping the third dimension to color.

Given the perceptual difficulties that exist with truly 3D representations being mapped into a 2D space, I’m not sure there would be significant value added by that functionality; perhaps a wire frame of a surface can help identify locations where a model might find local maxima, but there are other ways to represent the same information that are easier for end users to perceive visually.

Having contributed several packages to the Stata community, releasing a package on CRAN, and having Java binaries published to Maven archives, I too am unsure what specifically it is that you mean about research suffering because of a lack of custom code; in fact a research center at a US based university recently reached out to me about using one of the Java based packages I had developed as part of the code they were working on.

[Bjarte Aagnes]

Additional comment to #232:

Integration with ... Matlab

is possible via Stata Python integration, plugins (c, c++, java), or on Windows; automation (COM).

[Stephen Jenkins]

Please elaborate what you mean, giving some specific examples, for those of us who don't know R or Matlab intimately. As far as I understand an R "package" and a Matlab "toolbox" have analogues in Stata -- they are the programme or programmes (ado-files and help-files) that form a package of files. (Examples are the packages that one can download, for free, from SSC or the Stata Journal website.) R packages, e.g. those sitting on Cran, are typically free to download ... but many have to be user-downloaded (as in Stata). My googling re Matlab toolboxes suggests that one has to pay for some of them: see e.g. https://www.mathworks.com/help/thing...ox-access.html. So, another question is: what precisely do you mean by "custom"?

The one example you refer to is 3D graphics. I type -search 3D- in Stata and get a number of packages showing up, including contour plots and some 3D 'wire' diagrams. What functionality are you missing in particular? The ability to rotate 3D wire diagrams to show different perspectives, perhaps using a mouse? Or something else?

I don't want to miss out on any good stuff 'out there'; I am genuinely interested in learning about different ways of doing things. It's this that motivates my questions.

[Mario Ferri]

Integration with custom R and Matlab toolboxes and codes. Improvements in graphics, particularly 3Ds in order to reach the highest quality like R.
Currently research is done with custom codes and toolboxes. That's where stata suffers.

[ericmelse]

(SMCL?) CONTROL OF THE COLOR OF SOME OF THE CHARACTERS IN A STRING OF CHARACTERS

Stata has a good number of options to control the formatting of text, as well as the ability to use Greek letters etc., using SMCL tags, like:

sysuse auto , clear
twoway scatter price mpg , xtitle(Mileage ({bf:mpg}))

But, it is now not possible to control the color of individual characters of text within a particular string of characters, as to get a result like:

Mileage (mpg)

or

Mileage (mpg)

Would this be feasible in SMCL?
Perhaps something like:

xtitle(Mileage ({clr&blue:mpg}))

or

xtitle(Mileage ({clr&"255 0 255":mpg}))

[John Mullahy]

(I see this is my second consecutive wish; hope I'm not perceived as a troublemaker...)

Might the statistical functions normal(z) and invnormal(p) be modified to allow mean and standard deviation parameter arguments, e.g. normal(x,m,s), invnormal(p,m,s)? The normal density functions already allow this and I can't see why, in principle, this can't be done for the cumulative and inverse cumulative.

Thanks for considering.

[John Mullahy]

(Mac OS) In the Results window I occasionally like to zoom in (to magnify) or zoom out (to have a single image for a screen capture), often by several multiples.

In the Results window (as well as the Command window and Do-File editor) using "command plus +" and "command plus –" zoom in and out as indicated.

However I've been unable to find a keyboard shortcut or menu command that restores the default zoom. In web browsers (at least Firefox and Chrome) "command plus 0" restores the default, but "command plus 0" executed when the Results window is active activates the Command window, and executed when the Command window is active appears to do nothing.

Might v17 be able to build in the equivalent of the "command plus 0" keyboard shortcut to restore the default magnification?

[Allan Massie]

I would love to see new Do-file editor features (admittedly, I am missing the features I use in RStudio when forced to operate in R):

2) Shortcuts to (i) do/run the current chunk of code (either program definition or a section delineated by section headings, see above) and (ii) do/run from the top (currently we have Do to bottom, why not have the opposite -- which btw I find myself using much more often in R).

Jan, I like this idea. My workaround for "run from the top to cursor" (at least until there's an out-of-the-box shortcut) is to insert the line "assert 0" where I want it to stop

[Courtney Balentine]

It would be great to have more machine learning applications, like support vector machines, neural networks, random forest and other tree based algorithms. User written programs exist for these but would prefer to have as part of Stata

[wbuchanan]

Rick and Bruce,

I would take a different approach to this, which is used in a lot of other software development. Regardless of the source for the software (user contributed package), Stata should invest in a way to implement some type of versioning/retrieval similar to what exists in other languages. For example, R uses the Debian package specification which makes it possible to have multiple versions of the same package installed on the same machine that are localized to the package using it as a dependency. Python has similar structure (requirements.txt is the typical name for such a file). I think Java probably has the best solution for these types of issues with Maven; in addition to being able to bound dependent libraries with your library - and specify the precise version of the library to use - there are also options that allow you to compile the source code under the assumption that a library will already exist on the end user’s machine.

I would argue that this type of functionality would be more flexible than just sorting on the distribution date, because it would allow testing development releases while still allowing a stable release to be available to all. I know not everyone in the Stata community thinks the same way about solving this problem, but figured it could be useful to at least start up the dialog about this topic again.

[Richard Williams]

Bruce Weaver anybody who knew enough to do what you were suggesting would also know that the first match often isn’t the best match. My personal experience is that, when there are multiple matches, the ssc match is usually the best. People who don’t know that, though, may just go with the first thing they see, which is SJ. Therefore I think the default behavior should be ssc first. All results would still show but in a different order.

[Bruce Weaver]

Regarding #223, perhaps an ssc option could be added to search? E.g.,

search gologit2, sj
search gologit2, net
search gologit2, ssc

At the moment, only the first two of those work.