[Rich Goldstein]

1. -egen- should have options before() and after() (as -generate- does)

2. -meglm- should allow the identity link for each family (as -glm- does)

[Tiaga Falcao]

JanDitzen Networks Models are similar to Spatial Models, but with lots of endogeneity and node level variations. I am interested in node level preference and community level spill-over effect studies and was hoping that Stata could upgrade Spatial Models to cover models with Network-Like features or add a new section for Network Models in Stata 18.

[Maarten Buis]

Tiaga Falcao have you looked at www.nwcommands.org

[Chul-Kyoo Jung]

PLEASE PLEASE Do-file Auto Save !!!

[William Lisowski]

Chul-Kyoo Jung at #19 - On Stata 16.1 for Mac, the Preferences > General Preferences > Windows > Do-file Editor > Advanced shows me the following, which suggests two auto-save options are available. I can't comment on their effectiveness because I use an external editor for the bulk of my work.

[Ali Atia]

Small request for the do-file editor -- add the ability to use CTRL+# to switch between tabs without having to move the mouse and click.

[Leonardo Guizzetti]

Small request for the do-file editor -- add the ability to use CTRL+# to switch between tabs without having to move the mouse and click.

On Windows you can use Ctrl+Tab, and as I recall this also worked in Linux.

[Ali Atia]

Thanks -- this works; but the CTRL+# option would still be useful in cases where I want to move from the first tab to the fifth quickly without needing to scroll past tabs two through four.

[Leonardo Guizzetti]

Ah, I understand what you meant by this key combo now. I thought perhaps you were just in search of Ctrl+Tab. Scintilla appears to have this ability already, so maybe it would be an easy implementation.

[Clyde Schechter]

Re #19: Please, no! Or if you do put auto-save into the do-file editor, provide a way to turn it off.

In my work flow, I often build a new do-file by starting with an existing one and making modifications. That can sometimes by a long process, and I don't always save along the way because sometimes I am not sure until I am done, whether what I'm doing is going to be worth saving or not. The worst thing that could happen is to have auto-save kick in and clobber my original do-file during this process.

Following on #20, in Windows, the same approach offers an option to auto-save before do/run, though there is not an option to auto-save periodically. But I hope that auto-save before do/run, as an option, could be satisfactory to those who want an auto-save.

[Leonardo Guizzetti]

I agree with Clyde re auto-save. Some functionality exists there, but I really don't want it personally. I've been bitten a few times by auto-saving features from some alternative system in the past, and then I have to remember how to undo my work. For this reason too, I will also make use of git or scratch files if I want to be secure about which version is the official one, or which one is just a work in progress.

[Christopher Bratt]

Sergiy Radyakin I agree with #6: When we have several data frames open simultaneously, it would be great to be able to append one frame to another (that is, to merge data frames).

But while we are waiting for this to be implemented in Stata, take a look at -frameappend- See here: https://ideas.repec.org/c/boc/bocode/s458685.html

Works fine in my experience, even though it writes an odd warning to the user.

[Michelle Brown Ngogo]

I'd like to see conditional and marginal R2 for mixed effects models, as per Nakagawa & Schielzeth (2013) Meth Ecol Evol

[Tiaga Falcao]

Tiaga Falcao have you looked at www.nwcommands.org

I used nwcommands before. It delivers some network metrics only on small networks, and it is very slow on large networks (# of links > 20k) and its visualization is very outdated, beside there are other packages in R and Python that do better jobs. However, what I meant here was network "econometrics" and statistical analysis (different from metric calculation), which I don't know a commercial or academic solution for it yet Just heard few thing about Palantir but it's not available for public or academic use.

[Maarten Buis]

@Tiaga_Falcao fair enough, you want a regreression style analysis with inference. That is a tough problem in itself. But before StataCorp will implement that it will have to implement basic descriptive analysis of networks, and before that it will have to implement efficient ways of storing and manipulating networks in Stata. If StataCorp where to start right now, my guess would be it would be a two part release: the first release would contain the network storing, manipulation, and descriptive part. This is in part just because of the amount of time it would take to do this right (and fast, e.g. implementing algorithms for sparse matrices). In part it is also because that way they can get a feel for how users will use it and whether their design choices work for the users. The second release will contain regression style models for networks. My guess would be that getting the first release for network analysis done in time for stata 18 would extremely optimistic. So my guess for the earliest release with regression models for networks would be Stata 20. Now this is just my guess. It could also be that StataCorp has been working on this for a number of years now and it will be ready in 18.