[Andrew Lover]

Integrating Levenstein distance fuzzy string matching algorithms would be extremely useful. SAS had this function for a long time using COMPGED and it is much better at merging similar string data, such as names, than probabilistic matching.

Matchit

The matchit package (download from SSC) does exactly this, with Levenstein and a number of other algorithms.

[Igor Paploski]

Not sure if it's been mentioned before, but ridgeline plot ("joyplot") capabilities would be interesting.

More on: https://medium.com/the-stata-guide/c...s-dbe022e7264d

[Maarten Buis]

I don't understand the Igor Paploski 's request in #197. The link he included is about how to make that graph in Stata, so the capabilities are there.

[Igor Paploski]

Maarten Buis, perhaps I expressed myself poorly. There is a workaround for implementing the ridgeline plot, but it's not straightforward. It would be nice to see a dedicated command that streamlines the process - do you want to plot your data directly or do you want to smooth it? Do you want to normalize between-classes or not? How much do you want to squish the y-axis? Instead of asking the user to create the several ytop, y0, ynew, ytest, ytestnew that the workaround suggests and uses, this could be done behind the doors, facilitating user experience.

[Maarten Buis]

A lot of these choices that you want Stata to make for you are choices that are really situational, each graph will need different choices. So I don't think that this can be done "automagically", and if you tried everybody would be unhappy with the result. You could make those choices options in the command, but then you should ask yourself what the added value of that command really is. There might be some added value, and you could write that command easily enough. I am not that convinced about the usefulness of that graph itself and a potential command to create it, so I won't do it.

[Igor Paploski]

Well, I'm not asking for Stata to make those choices for me nor for it to be done automagically. I'm just stating that having a command that lays those options to the user in a friendlier way (instead of making the user create several intermediate variables) would facilitate user experience. Last, I wish I had the programming knowledge to create that command, but alas - another reason for this to be posted on a wish list topic. I definitely should put some time into learning Stata programming.

Cheers

[Nick Cox]

#197 onwards

My own biases include

1. These plots are enjoyed partly for the wrong reasons, as reminiscent of rolling landscapes or even of loosely similar human anatomy.

2. Some discussions seem to imply that some series occluding other series is, if not a feature, then not a problem. I can't see that.

But I guess many of us would like to experiment if and when Asjad Naqvi offers a command with moderate generality rather than a worked example.

The idea of superimposing or juxtaposing estimated density or other functional traces is fine and standard and already available with just a little bit of work.

[Lars Pete]

I wish better integration with LaTeX and ability to create publishworthy tables using simple commands (not using frmttable and matrices). Tabout simply doesn't work and gives too many errors.

[Stephen Jenkins]

#203: Lars, have you checked out Ben Jann's -estout- suite (on SSC; with extensive webpage documentation pages as well), or the new -table- command in Stata 17 including -etable- in a recent update? Put differently, it would help if you were to say more precisely what sort of tables you were hoping to create while citing problems/omissions with existing functionality. As ever, explicit examples help too.

[Fei Wang]

The -gsem- seems a little weak in allowing errors from different equations of generalized responses to be correlated. One may construct such cross-equation correlations through latent variables, but coefficients need to be rescaled, as in here. The user-written command -cmp- is good enough, but I'm still looking forward to an official command that handles general situations in the joint estimation of equations.

[Pantelis Kazakis]

If not possible already, allow the loading of different color themes for the do-file editor. There are already plenty of themes out there and it would be great if Stata could do this too. RStudio does this already.

[Jean-Michel Galarneau]

Stata to highlight, in the do-file, where an error occurred when a mistake is made in the syntax... not always obvious where the mistake occurred when running large chunks of script.

[JanDitzen]

Stata to highlight, in the do-file, where an error occurred when a mistake is made in the syntax... not always obvious where the mistake occurred when running large chunks of script.

the
Like in Matlab or RStudio? I think this is a nice idea, but to the best of my knowledge the do-file editor is not directly connected to Stata. It saves a temporary file which is then run. I believe it is technical possible, but hard to achieve?

In the meantime, have you tried set trace on? It doesn't tell you the line where the error is, but it shows the command line at least.

[Adam Streff]

It would be great if there could be an option with the month() function to automatically create month labels instead of always having to create a value label ourselves (i.e. local mon 1 "Jan" 2 "Feb" 3 "Mar" ...)

[Leonardo Guizzetti]

It would be great if there could be an option with the month() function to automatically create month labels instead of always having to create a value label ourselves (i.e. local mon 1 "Jan" 2 "Feb" 3 "Mar" ...)

Have you tried to use date time formats? That's their purpose.