example.org

set up Org Mode (one time)

we have a hypothesis relating the density of a US state with the life expectancy in that state. let’s write some code to explore this.

emacs, org mode setup

to allow evaluating R code, evaluate this source block by putting point inside and typing [C-c C-c] (control C followed by control C) and giving permission for it to be evaluated.

(org-babel-do-load-languages
 'org-babel-load-languages
 '((emacs-lisp . t) (R . t)))

(normally, you would do this by customizing org-babel-load-languages.)

then, to allow editing R code, evaluate this source block, etc.

(require 'ess-r-mode)

(this is required because i’m running emacs -Q, so i don’t have my normal customizations.)

first R source block – lookup something

let’s evaluate this

state.abb[match("Wyoming", state.name)]

but, we might need to do that many times. let’s write a subroutine to do that, passing the table from which the return value should be taken as a second argument.

lookup <- function(state, where) {
  where[match(state, where)] # XXX
}

lookup("Wyoming", state.abb)

we seem a bit low on results. what’s wrong?

find some state statistics

state.x77[,"HS Grad"][match("Wyoming", state.name)]

but, maybe we could re-use the lookup function? using :noweb?

add a name to the source block that held the lookup function.

lookup <- function(state, where) {
  where[match(state, state.name)]
}

okay, now we can include it with :noweb

<<lookup>>

lookup("Wyoming", state.x77[,"HS Grad"])

now, do a graph

for our project, we’re going to be accessing various state statistics, so we realize we’d like also to have the lookup of state statistics as a common routine.

re-use

we’ll need some utilty code to return columns of state.x77 (which we can include with :noweb).

state_stat <- function(which) {
  state.x77[,which]
}

and, make use of that, see if it works.

<<state_stat>>
<<lookup>>

lookup("Wyoming", state_stat("HS Grad"))

now, routines to compute the specific stats we want

we take the calculations of the relevant statistics out of the block that actually does the graphing. to pass the results, we can’t use :noweb. we can, however, use :var, which allows the results of one source block to be used as input to another source block.

a simple routine to produce the life expectancies of states. note that :results is set to table. (adding drawer means we can hide it away easier, to keep the visual buffer less cluttered.)

<<state_stat>>
state_stat("Life Exp")

now, what is the density? population=/=area, i guess.

<<state_stat>>
state_stat("Population") %/% state_stat("Area") # XXX

finally, the graphing code itself.

and, now, write the graph routine. reminder: we aren’t accessing density and expectancy via :noweb, but, rather, via :var (but, we still use :noweb).

i hope it works!

<<lookup>>  # note: syntax checker doesn't understand =:noweb= references

plot(density, expectancy,
     panel.first = lines(stats::lowess(density, expectancy), lty = "dashed"),
     pch = ".", cex = 1.2, col = "blue")
text(density, expectancy,  lookup(rownames(state.x77), state.abb),
     pos=1, cex=1, col="red")

(open the resulting file with [C-c C-o].)

now, let’s export and see our marvelous work!