iebaltab rewrite kb

[kbjarkefur]

Please provide comments in this PR

[bbdaniels]

make option onerow the default when the number of observations is not changing

I vote no for backward compatibility. We have broken backward compatibility in this version for some cases where it is a really good case for it. I do not think this is as important.

My vote would be to do it now. If we're already breaking backward compatibility, better to make all changes we want to that will break it now. That's a weak preference, though. @bbdaniels , what do you think?

I am always in favor of breaking compatibility for big functionality upgrades at major versions! But I don't think this is supposed to be a major version -- primarily a rewrite so that we can do that in the future? So in this case I would say keep things backward compatible and don't have a major version, then let's carry over a list of compatibility-breaking functionality and syntax changes for the next version.

[kbjarkefur]

add option to make "Total" the first column in the table

How important? Can this be pushed to a later version? I am keen to get this version out, then we can add new features that we want.

Yeah, this can come later. It's just a bit counter intuitive, but nothing major. Moved it to issue #298

Fixed in 851a9c2

[kbjarkefur]

make option onerow the default when the number of observations is not changing

Wont implement due to this argument made on Slack:

I am still not convinced this is the best way to got. Lets say all my columns has the same N. With the new default I would only have my Ns on a single row at the bottom. If then something changes to my data so one data point is missing. Then I would not get an error or a warning, instead the structure of my table would tacitly change. Is it only me who sees that as concerning or confusion at best. I still vote for keeping it as it is. As in the current default, the layout of the table does not change as the data change. And by adding the option onerow you add a requirement that the data should be in a certain way (i.e. N the same in a column across all rows). For those that know their data will always meet this condition they can run iebaltab with onerow always there, just like all of us who know we want to replace the already existing file run iebaltab with replace.

[luizaandrade]

From looking at the compiled latex file and comparing with the current SSC version, all the estimates are consistent, but the display still has some issues.
v6.4.pdf
v6.3.pdf

• significance stars are not being exported, even when the stars option is explicitly called
• can we leave the double bars at the bottom of the table?
• can we also leave the extra spacing between variables? I find that easier to read
• SEs are not in parenthesis
• can we add a horizontal line before number of obs if using one row?
• can we add an option to not show the number of obs for the pairwise tests?

[kbjarkefur]

SEs are not in parenthesis

Fixed in 3cef80d

[kbjarkefur]

significance stars are not being exported, even when the stars option is explicitly called

There is no option called stars, you should have received an error in the new version with that option. Anyways, there was a bug suppressing the stars in the tex tables, that is fixed in 3657647

[kbjarkefur]

can we leave the double bars at the bottom of the table?
can we add a horizontal line before number of obs if using one row?

fixed in c6418bf - but lests discuss if there should be one or two single lines when both f-test and onerow are used

[kbjarkefur]

can we also leave the extra spacing between variables? I find that easier to read

fixed in 0c5b756 - I also took the liberty to deprecate the option texvspace(string) that customize that space, but for every row. The whole point of returning results in matrices is for users to customize their tex tables outside iebaltab.

[kbjarkefur]

can we add an option to not show the number of obs for the pairwise tests?

Suggest adding to an issue to get this version out

[luizaandrade]

significance stars are not being exported, even when the stars option is explicitly called

There is no option called stars, you should have received an error in the new version with that option. Anyways, there was a bug suppressing the stars in the tex tables, that is fixed in 3657647

Sorry, I meant the starlevels() option