[R] How good is R at making publication quality tables?

[Douglas Bates]

On Wed, Mar 17, 2010 at 10:25 AM, Erik Iverson <eriki at ccbr.umn.edu> wrote:
>
>
> Paul Miller wrote:
>>
>> Hello Everyone,
>>
>> I have just started learning R and am in the process of figuring out
>> what it can and can't do. I must say I am very impressed with R so
>> far and am amazed that something this good can actually be free.
>>
>> Recently, I finished reading R for SAS and SPSS Users and have begun
>> reading SAS and R and Data Manipulation with R. Based on what I've
>> read in these books and elsewhere, I get the impression that R is
>> very good at drawing high quality graphs but maybe not so good at
>> creating nice looking tables of the sort I'm used to getting through
>> SAS ODS.
>
> You're really only limited by your imagination here.  I have written several
> custom table functions to output LaTeX, but you can output whatever you like
> (HTML, plain-text, org-mode files...), you're in complete control with R.
>
> I can second the Hmisc package though.  I often use a combination of
> summary.formula and the latex function to output really nice looking tables
> that get put into a long PDF report for a study.
>
> I can say that both of these functions, summary.formula and latex, in Hmisc
> have a LOT of arguments, and almost every time I said "I wish it looked a
> little different", there was an option to control it.
>
> Specifically, I found the options:
>
> exclude1, long, longtable, combine, test,
>
> do be very useful.  I often make tables by some treatment group, so all
> these are using method = "reverse" to accomplish that.
>
> And if you don't like the output, latex.summary.formula.reverse is a good
> function to make your own version of, to output exactly what you want.  I
> have a local copy that augments the tables that contain unadjusted p-values
> with adjusted p-values from a model.
>
> But apart from Hmisc, just realize that with R you have a nice programming
> language to produce any type of output you want, you're not limited to what
> someone else gave you.

Another option to consider is the xtable package.