Wish list for Stata 14

[Richard Williams]

As far as I can tell, nobody else has started a thread like this on the new board, so I figured I would go first. Having all the wishes compiled in a single thread might be helpful.

My wishes include:

Command specific help for margins (or else some FAQs). People are always asking why margins does or doesn't do this or that after some command. Panel data techniques seem especially problematic. The answer is usually that some option is not appropriate given the technique used. Stats geniuses might already realize this, but for the rest of us some sort of at least brief explanation would help.

Better support for margins with multiple outcome commands. After commands like ologit and mlogit, you have to run a separate margins command for each outcome of the dependent variable. I'd like to have margins do it with one command.

More powerful factor variables. I'd like to be able to specify more functions of independent variables, e.g. the log of a variable, the cube root, whatever. Then have margins realize that functions of the same variable are related to each other, e.g. the value of X is related to the value of the square root of X.

[Clyde Schechter]

I would find it helpful if there were a little more consistency in the names of options across commands that do similar things. For example, when I -export delimited- I control whether or not the first row of my output file contains the variable names using the option -novarnames-. But when I -export excel-, the same functionality is controlled by -firstrow(variables)-.

Now the aforementioned options do have somewhat different functionality, in that with -import excel- we have the choice of putting the variable labels in the output rather than the names. But that functionality could also be useful in -export delimited-. And in any case, probably just having a -varnames- option specification in both commands would be simpler.

There are scattered other commands across Stata where commands accomplish highly similar functionality with options that are differently named.

[Clyde Schechter]

Sorry, hit Post Reply when I was aiming for Preview.

Another example of uniformity would be if we could just use -,eform- to get exponentiated coefficients in all regression model output (perhaps still retaining the more specific options -or-, -irr- that do this for specific models as well).

[Elan Cohen]

Hopefully this is an easy implementation, but I would love a sepby option for browse just as there is for list. My first thoughts were either a darker/bolder line for separation, or alternate background shading.

[Richard Williams]

Sorry, hit Post Reply when I was aiming for Preview. Another example of uniformity would be if we could just use -,eform- to get exponentiated coefficients in all regression model output (perhaps still retaining the more specific options -or-, -irr- that do this for specific models as well).

I like the eform option too. The main counter-argument may that forcing you to use the "right" names also forces you to understand why the same transformation is considered an odds ratio in one statistical technique but an incident rate ratio in another.

[Timothy Mak]

On the topic of browse, I wish that Stata had retained the old (Stata 11 or before?) behaviour where executing browse pauses execution of codes. I used to put in browse statements in my codes for debugging, which I find very convenient. I can't do this anymore.

[Daniel Bela]

On the topic of browse, I wish that Stata had retained the old (Stata 11 or before?) behaviour where executing browse pauses execution of codes. I used to put in browse statements in my codes for debugging, which I find very convenient. I can't do this anymore.

Did you ever try pause? I find it very convenient, as it helps debugging much more than the old browse behaviour (i.e. inside of loops or between preserve and restore). Plus: it can be simply disabled once testing has finished using pause off.

pause on
sysuse auto
pause Check if dataset has been loaded correctly
tabulate make in 1/5
*...

Regards
Bela

[wbuchanan]

I would like to see more extensive documentation for the graphics back end to make it a bit easier for end-users to add to the visualization capabilities with greater ease. Another option that could be helpful is to develop interfaces for building interactive graphics using existing libraries (e.g., d3) to avoid reinventing the wheel.

[Richard Williams]

I want the opposite of pause. Once a job is running, I want to be able to -set more off-. All too often I forget to do that before I start the job and then I have to either kill the job or keep clicking on more.

[Jeph Herrin]

Some version of the -case- (also called -switch- or -select-) command that is available in other programming languages. I usually accomplish the same thing with either a series of -if- statements or set of nested -cond()- calls; neither is very easy to read.

[Timothy Mak]

Thanks Bela. I didn't know about pause.

[Robert Wells]

I want the opposite of pause. Once a job is running, I want to be able to -set more off-. All too often I forget to do that before I start the job and then I have to either kill the job or keep clicking on more.

Not a perfect solution Richard, but I find that -set more off, perm- gets around this problem as it set more off. i find it much more useful to have more off indefintely than the alternative. still i would like an on the fly set more on and off too

[Dave Airey]

Stata 13 was the first version that I scratched my head about what I might want added. But here are some quick thoughts:

• nonlinear mixed effects models (useful in Pharma)
• integration with markdown
• after R "rms" and JMP 11, regression modeling techniques to help build predictive models including Ridge, Lasso, adaptive Lasso, and Elastic Net
• opacity choice in graphics

[David Lucido]

I feel that Stata's greatest limitation is the lack of any camera-ready output capabilities outside of graphics. I can't even count the number of table programs that produce TeX code which then has to be re-processed. I would hope Stata could produce an output system that allows for fonts, proportional spacing, styles etc. and which can create PDF, RTF or any other standard format. I am thinking of something like the Output Delivery System available in SAS.

[Lance Erickson]

Originally posted by Richard Williams View Post
I want the opposite of pause. Once a job is running, I want to be able to -set more off-. All too often I forget to do that before I start the job and then I have to either kill the job or keep clicking on more.

Also not the best solution, but on a slower comupter, a double tap on the space bar will clear all the -more-s and let Stata run through all of the output. On a faster computer I have to hold the space bar down to get the same outcome. I've also noticed that if I already have a couple of spaces in the command line then I get the same result as if -set more- was -off-.