iebaltab: layout updates

[luizaandrade]

I have a project at the Lab that needs to create a balance table with two panels. So I figured I'd implement this in iebaltab. My idea was to have a similar workflow as in estout by adding the following options:

• Fragment
• prehead & posthead to customize headers (and thus solve [iebaltab]: headers for groups of variables #236)
• prefoot & postfoot to customize footers
• Add option to suppress SE below mean in iebaltab #147
• iebaltab: add numbers to all columns #144 (if not yet implemented)

anything else @kbjarkefur and @bbdaniels think should be included here?

[kbjarkefur]

Would this only be for LaTeX? If not, then I think we should be specific what we want to do for LaTeX only and what we want to do for Excel and browse.

[kbjarkefur]

I have a strong preference to no add separate options that specifically suppress certain stats. Instead, I think this is intended to be a part of the stats() option, see this issue: #293. A big part of my implementation in the re-write was to move away with the ad-hoc approach of separate options for specific layout options.

So that stats(desc(mean/se)) would be how you configure what is currently the default (happy to discuss whether that should be the default), and to drop the SE below mean you would do stats(desc(mean)).

[luizaandrade]

Yes, this would be only for LaTeX. I agree on the stats() option. I don't think the suggested option from #147 should be the default. If this is already implemented as stats(desc(mean)), I think we can mark that issue as solved.

[kbjarkefur]

I limited the scope of my big re-write to gather all already existing separate stats display options into stats(), but I did not add any new functionality to it. I had to limit the scope somehow to ever get that version out the door.

So instead of the stdev option you can now do stats(desc(sd)), but just as before the only option you have is to change what is displayed on the second row, not changing that there is no second row or that it is empty.

I think the solution is to allow users to set one or many stats in all of stats(desc() pair() f() feq()) and then the tables adjust to have as many rows needed for each of them. I have more thinking on the syntax for setting multiple options, but we should probably discuss that in the issue for that option.

[bbdaniels]

I like this solution -- I should certainly be able to request stats( [...] pair(beta f p))!