[R] plotting on a reverse log scale

[Gabor Grothendieck]

There is an enhanced pretty function called nice in the Epi package
that also works with log data.  Not sure if this could lead to any 
simplifications in this problem?

On 7/6/05, Marc Schwartz <MSchwartz at mn.rr.com> wrote:
> I thought that I would take a stab at this. I should note however that
> my solution is heavily biased by the first log scale plot that Michael
> referenced below. To wit, my x axis has major ticks at powers of 10 and
> minor ticks at 1/10 of each major tick interval.
> 
> Thus, here is my approach:
> 
> # Set the max range to an integer power of 10 as may be required
> max.pow <- 5
> 
> # Set the data as others have done below
> set.seed(1)
> x <- runif(500, 1, 10^max.pow)
> d <- density(x, from = 1, to = 10^max.pow)
> 
> # Now do the density plot, leaving the x axis blank
> # Note that by doing a rev() on 'xlim' it reverses the min/max on the
> # x axis, so you don't have to worry about adjusting things.
> plot(d$y ~ d$x, xlim = rev(c(1, 10^max.pow)), log = "x",
>     xaxt = "n", type = "l",
>     xlab = "2000 - Year (log scale)")
> 
> # Set the axis major tick marks
> axis.at <- 10 ^ c(0:max.pow)
> 
> # Draw the major tick marks and label them using plotmath
> axis(1, at = axis.at, tcl = -1,
>     labels = parse(text = paste("10^", 0:max.pow, sep = "")))
> 
> # Now do the minor ticks, at 1/10 of each power of 10 interval
> axis(1, at = 1:10 * rep(axis.at[-1] / 10, each = 10),
>     tcl = -0.5, labels = FALSE)
> 
> # Do the rug plot
> rug(x)
> 
> 
> Not sure if this is helpful here, but thought I would post it for
> review/critique. The 'max.pow' constant can be explicitly adjusted or
> can be calculated automatically based upon input year ranges.
> 
> Best regards,
> 
> Marc Schwartz
> 
> 
> On Wed, 2005-07-06 at 17:31 -0400, Gabor Grothendieck wrote:
> > Not sure if I am missing something essential here but it would
> > seem as simple as:
> >
> > # data
> > set.seed(1)
> > x <- runif(500, 1500, 1990)
> >
> > # plot
> > d <- density(x, from = 1500, to = 1990)
> > plot(d$y ~ d$x, log = "x")
> > rug(x)
> > axis(1, seq(1500, 1990, 10), FALSE, tcl = -0.3)
> >
> >
> > On 7/6/05, Michael Friendly <friendly at yorku.ca> wrote:
> > > Thanks Duncan,
> > > That is almost exactly what I want, except I want time to
> > > go in the normal order, not backwards, so:
> > >
> > > # plot on reverse log scale
> > > years1500 <- runif(500, 1500, 1990)  # some fake data
> > > x <- -log(2000-years1500)
> > > from <- -log(2000-1990)
> > > to <- -log(2000-1500)
> > > plot(density(x, from=from, to=to), axes=F)
> > > rug(x)
> > >
> > > labels <- pretty(years1500)
> > > labels <- labels[labels<2000]
> > > axis(1, labels, at=-log(2000-labels))
> > >
> > > minorticks <- pretty(years1500, n=20)
> > > minorticks <- minorticks[minorticks<2000]
> > > axis(1, labels=FALSE, at=-log(2000-minorticks), tcl=-0.25)
> > >
> > > axis(2)
> > > box()
> > >
> > > -Michael
> > >
> > > Duncan Murdoch wrote:
> > >
> > > > On 7/6/2005 3:36 PM, Michael Friendly wrote:
> > > >
> > > >> I'd like to do some plots of historical event data on a reverse log
> > > >> scale, started, say at the year 2000 and going
> > > >> backwards in time, with tick marks spaced according to
> > > >> log(2000-year).  For example, see:
> > > >>
> > > >> http://euclid.psych.yorku.ca/SCS/Gallery/images/log-timeline.gif
> > > >>
> > > >> As an example, I'd like to create a density plot of such data with the
> > > >> horizontal axis reverse-logged,
> > > >> a transformation of this image:
> > > >> http://euclid.psych.yorku.ca/SCS/Gallery/milestone/Test/mileyears1.gif
> > > >>
> > > >> Some initial code to do a standard density plot looks like this:
> > > >>
> > > >> mileyears <- read.csv("mileyears3.csv", skip=1,
> > > >> col.names=c("key","year","where","add","junk"))
> > > >> mileyears <- mileyears[,2:4]
> > > >>
> > > >> years <- mileyears$year
> > > >> years1500 <- years[years>1500]
> > > >> dens <- density(years1500, from=1500, to=1990)
> > > >> plot(dens)
> > > >> rug(years1500)
> > > >>
> > > >> I could calculate log(2000-year), but I'm not sure how to do the
> > > >> plotting, do some minor tick marks
> > > >> and label the major ones, say at 100 year intervals.
> > > >
> > > >
> > > > I think you'll have to do everything explicitly.  That is, something
> > > > like this:
> > > >
> > > > years1500 <- runif(500, 1500, 1990)  # some fake data
> > > > x <- log(2000-years1500)
> > > > from <- log(2000-1990)
> > > > to <- log(2000-1500)
> > > > plot(density(x, from=from, to=to), axes=F)
> > > > rug(x)
> > > >
> > > > labels <- pretty(years1500)
> > > > labels <- labels[labels<2000]
> > > > axis(1, labels, at=log(2000-labels))
> > > >
> > > > minorticks <- pretty(years1500, n=20)
> > > > minorticks <- minorticks[minorticks<2000]
> > > > axis(1, labels=FALSE, at=log(2000-minorticks), tcl=-0.25)
> > > >
> > > > axis(2)
> > > > box()
> > > >
> 
>