Design notes

These are some names I played with:

aod :: arangment of data
yat :: yet another table
drc :: data in rows and columns
rcd :: rows and columns data
mft :: my friend's table
mftf :: my friend's table functions
rcd :: rows and columns data
tbl :: table
tay :: tables 
able :: a short table
tapi :: table API
apit :: API for tables
yast :: yet another summary table
yadt :: yet another descriptive table
yalmt :: yet another lm table
yalmert :: yet another lmer table
dst :: descriptive statistics table

But then I realized there can't be one function to do everything. For example, the code to generate a table of descriptive statistics would be quite different from the code to extract coef(summary(fit)) and put it into a table.

So this is what I have

• tods or toss is a function to create a table of descriptive statisics
• tolm creates a table from an lm model
• etc.

Table of summary statistics

The formula interface would look like this:

toss(margin(group) + group + p(group) ~
	   mediqr(age) +
	   npc(sex),
	 data = dat)

The functions mediqr (for median and IQR) and npc (for number and percent) are functions that calculate the desired summaries. These could take up multiple rows in the table like this:

	Total	Group A	Group B	P-value
Age, median (IQR)
Sex, n (%)
<tab>Men
<tab>Women

There can be some formating in functions like mediqr with the first argument being something like didits.

mediqr

This function and others like it have a fmt argument that is similar in spirit to the (first) fmt argument of sprintf.

mediqr(age, fmt = "%f (%f, %f)")

The fmt string could be extended to encode stying issues like alignment and border.

I guess these summary functions can return an object of descriptive_statistic (or shorter name) and since the object has an fmt element the print method can look at this element?

Table headers and caption

I can imagine something like this:

tods(group ~ mediqr(age, th = "Age, years"), 
     data = dat,
	 caption = "Summary table",
	 header = ???

I don't know good syntax for the column headers. In HTML5 these are <tr> tags with a scope (I think).

What about this:

header = c(tr(th("Name", rowspan = 2),
              th("ID", rowspan = 2),
			  th("Membership", colspan = 2),
			  th("Balance", rowspan = 2)),
		   tr(th("Joined"),
		      th("Cancelled")))

Revisiting the package notes (July 27, 2025)

I forgot that yat is already useful. It creates a data frame of summary statistics. If yat is for the descriptives should there be something like yatlm or should yat be yatd (yet another table of descriptives).

The package can be yatp for "yet another table package".

Or should it be yat1 for yat another table 1.?

Formatting

Currently the data frame holding the table content has formatted values and the underlying values are lost. Is this OK? I am tempted to try to retain the numerical summary statistics but that is probably more overhead than I need. I could see how far I can go with the formatted values. (I could have a function output the raw values but a value like "3.8 (1.7)" would have two raw values. Is that a list?

If I have toOrg(yatd(...)) then toOrg can sort out whether I have formatted cells or raw values. This tells me that the basic use would be to store formatted values in the cells.