[FernandoRios]

Something simple, but perhaps useful.
It would be great if margins could store the results in e() without having to type post.
something like:
margins, dydx(x1 x2 x3) estore(m1)
That way, we do not "lose" the previous model, and still store the results for exploration

[Julian Reif]

The recode() command has a test option that displays a warning if the recoding rules overlap. It would be nice to have an option that generates an error rather than just a warning.

[Rich Goldstein]

-ologit- by default uses the lowest value of the outcome variable as its comparator; I would like an option added so that I could choose whether I wanted the lowest or the highest value; yes, I can "fake" this by reversing the order, but this does not quite match what other software does where the default in the other software is to choose the highest value (the cutpoints will have opposite signs using the Stata fix as compared to what the other software is doing) - I realize that this is not generally a big deal but when working with others using different software, it would be very convenieint to easily show them the results match by just changing the comparator (without having to explain the change in signs)

[Jesse Wursten]

Stata string matrix. One use case is to use a string matrix as loop-up table.

Can't you use frames for this now?

[Gordon Fick]

I think that Stata should have a command for generalized additive models [ GAM ]. gamfit.exe is very buggy and I believe it runs only on MSWindows machines. A built-in command would be so much better.

[Bjarte Aagnes]

Stata string matrix. One use case is to use a string matrix as loop-up table.

[Daniel Feenberg]

The Methods and Formulas section of [SVY] Subpopulation Estimation suggests that estimating subpopulation characteristics by simply subsetting the data and applying the sampling weights is not appropriate.

Does -statsby- intend to rule out probability weights when estimating the same regression across several countries? I suppose countries are each subsets of the world, but isn't it ok to have separate regressions by country since variations in the probability of being sampled won't move an observation from one country to another? Isn't -statsby- presuming the weights arise from a particular sample design, when other designs are also possible?

[William Lisowski]

#595 - Stata developers are aware of this problem introduced by macOS Ventura and have said it will be corrected in a forthcoming update to Stata 17.

https://www.statalist.org/forums/for...-and-ado-files

[Haoran Liu]

Preview .do and .ado files on MacOS by clicking space.

[Gorkem Aksaray]

Chen Samulsion for #565 Would instead a syntax like collapse, force frame(new_frame_name) work better? My idea would be that that would make a new frame (new_frame_name) with the collapsed data in it and change to that frame.

Maarten Buis regarding #575

One alternative is to use the user-written prefix command frapply available from SSC:

ssc install frapply, replace
sysuse auto
frapply, into(newframe, replace): contract rep78 foreign |> sort foreign rep78 |> list

frapply is flexible; one can apply commands (such as collapse) to any frame, replace or change the current frame to the new frame, and daisy-chain commands. See my discussion here: https://www.statalist.org/forums/for...d-s-to-a-frame.

[John Williamss]

Add code block support for magic command %stata in pystata.
This is crucial for the complete integration of python with Stata, i.e., running a big python and stata code inside python IDE (like Spyder or Jupyter) with pystata integration.

Right now, the only workaround is to use the %%stata or stata.run command for multiline support (e.g., for writing a Stata for block in python IDE), but that should be executed separately. Suppose we want to loop over a variable in python and each time run a multiline block/command in PyStata. Right now, there is no way to do that. For example, there is no way to make this code work:

 %loading pystata as usual
import stata_setup
stata_setup.config("C:/Program Files/Stata17", "mp")
from pystata import stata
from sfi import Macro  

for ii in [1,2]:  
 Macro.setLocal('kk', str(ii))
  %stata   forvalues jj=1/`kk'
  %stata  {  
  %stata  disp `jj'
  %stata  }

[Clyde Schechter]

Re: #590 and #591. Anything you can do with -statsby-, and more, can also be done with -runby-, by Robert Picard and me, available from SSC. -runby- actually imposes no restrictions at all on the commands it processes. Any commands that can run standalone in Stata, can be run group-wise using -runby-, and in large data sets much faster. The only major limitation with -runby- is that the commands it runs cannot access local macros defined in the calling program (there is no problem with macros defined within the commands being executed by -runby-), so any such information needed must be passed in some other way.

So, if you are sure that what you want to do with -pweights- can still produce correct answers when carried out groupwise, then you can do it with -runby-. Of course, you do this at your own peril if your judgment about the suitability of groupwise calculation is ill-founded.

[William Lisowski]

The Methods and Formulas section of [SVY] Subpopulation Estimation suggests that estimating subpopulation characteristics by simply subsetting the data and applying the sampling weights is not appropriate.

[Daniel Feenberg]

Allow -statsby- to work with pweights. https://www.stata.com/manuals/dstatsby.pdf says

All weight types supported by command are allowed except pweights;

I wonder why -statsby- cares, other than a belief that use of pweights should be discouraged.

[Joseph Coveney]

I wonder if

cond(x,a(z),b(z))

evaluates both a(z) and b(z) before evaluating x, rather than using x to choose which one of them to evaluate for the current observation?

cond(x, a, b) evaluates b, then a and then x irrespective of whether x is true or false.

version 17.0

clear *

log close _all
log using condeval.txt, text nomsg name(lo)

local seed 1685397315
local 0 = "runiform() < 0"
local 1 = "runiform() > 0"
local show display in smcl as text

set seed `seed'
`show' `0'
`show' runiform()
`show' runiform()

set seed `seed'
`show' `1'
`show' runiform()
`show' runiform()

set seed `seed'
`show' -1
`show' runiform()
`show' runiform()

set seed `seed'
`show' -1
* `show' runiform()
`show' runiform()

set seed `seed'
`show' cond(`0', runiform(), runiform())
`show' runiform()

set seed `seed'
`show' cond(`1', runiform(), runiform())
`show' runiform()

set seed `seed'
`show' cond(`1', runiform(), -1)
`show' runiform()

set seed `seed'
`show' cond(`1', -1, runiform())
`show' runiform()

quietly log close lo

exit

I suppose that you could use the same diagnostic approach if you're interested in how the function behaves with its optional (three-valued logic condition) fourth argument.