[R] plotmath question

[Deepayan Sarkar]

On Friday 18 March 2005 15:41, Berton Gunter wrote:
> R listers:
>
> I have been foiled by plotmath!
>
> (in R 2.01,Windows 2000)
>
> The task: Plot a normal density and label the ticks as mu - 3 sigma, mu - 2
> sigma, ...., mu + 3 sigma, where the mu's and sigmas appear as Greek
> symbols, of course.
>
> The following code does this:
>
> x<-seq(-3,to=3,by=.01)
> y<-dnorm(x)
> plot(x,y,type='h',col='lightblue',axes=FALSE)
> lines(x,y,col='darkblue')
> axis(2)
> for(i in seq(-3,to=3))
>  axis(1,at=i, lab=switch(sign(i)+2,
>    eval(substitute(expression(mu-j*sigma),list(j=-i))),
>    expression(mu),
>    eval(substitute(expression(mu+j*sigma),list(j=i)))))
> box()
>
> However, I think the code in the for loop is ugly and probably means that
> I'm doing it wrong. In particular:
>
> 1) Is there a neat way to use one axis() call and a vector (of
> expressions?) for the lab=argument?

Yes, expression objects can be vectors. e.g.:

## use switch as above for better formatting
lab = do.call(expression, 
              lapply(-3:3, function(i) { 
                   bquote(mu + .(i) * sigma) 
              } ))
axis(1, at = -3:3, lab = lab)


> 2) The plotmath Help state that expressions can be used for axis labels, so
> I would have expected the above to work without the eval()call  -- but it
> does not. Would someone kindly explain to me why not -- i.e., what I have
> misunderstood. That is, to be clear, why does the following not work:
>
> for(i in seq(-3,to=3))
>  axis(1,at=i, lab=switch(sign(i)+2,
>    substitute(expression(mu-j*sigma),list(j=i)),
>    expression(mu),
>    substitute(expression(mu+j*sigma),list(j=i))))


> is.expression(substitute(expression(mu-j*sigma),list(j=1)))
[1] FALSE
> is.expression(eval(substitute(expression(mu-j*sigma),list(j=1))))
[1] TRUE

?substitute says

Value:

     The 'mode' of the result is generally '"call"' ...

which evidently have to be evaluated.

Hth,

Deepayan