R help - Mar 2006 - trellis graph question

R2.2, WinXP:

I am using xyplot( ) to generate time plots of plasma concentration data.
The following is an edited version of my code:

xyplot(log.conc ~ time| group, data = foo,
	groups = subject,
        panel = function(x, y, panel.number, ...)
	{
	 panel.superpose(x, y,
	                 subscripts = TRUE, 
                         groups = foo$group,
	                 type = 'l', col = "black")
	 panel.loess(x,y,span = .66, col = "red", lwd = 3)
         panel.grid(h=-1, v=-1, lwd = 1)
	}
       )

For each dose group, I want to plot all subjects within the group and
superimpose a loess smoother so that an overall trend can be made
apparent.  The code above does this, but has the annoying tendency to
"connect" the first-last points for (say) subject 1 with the
last-first
points of (say) subject 2.....

This annoyance is something that is tolerable for exploratory graphs
but not for publication-quality graphs... Is there a "fix" to this
"connect-the-wrong-points" problem?


Much thanks and gratitude in advance,

    Greg

On 3/28/06, Greg Tarpinian <sasprog474474 at yahoo.com>
wrote:
> R2.2, WinXP:
>
> I am using xyplot( ) to generate time plots of plasma concentration data.
> The following is an edited version of my code:
>
> xyplot(log.conc ~ time| group, data = foo,
> 	groups = subject,
>         panel = function(x, y, panel.number, ...)
> 	{
> 	 panel.superpose(x, y,
> 	                 subscripts = TRUE,
>                          groups = foo$group,
> 	                 type = 'l', col = "black")
> 	 panel.loess(x,y,span = .66, col = "red", lwd = 3)
>          panel.grid(h=-1, v=-1, lwd = 1)
> 	}
>        )
>
> For each dose group, I want to plot all subjects within the group and
> superimpose a loess smoother so that an overall trend can be made
> apparent.  The code above does this, but has the annoying tendency to
> "connect" the first-last points for (say) subject 1 with the
last-first
> points of (say) subject 2.....
>
> This annoyance is something that is tolerable for exploratory graphs
> but not for publication-quality graphs... Is there a "fix" to
this
> "connect-the-wrong-points" problem?
I'm confused by your panel function, but from your description, the
following should do what you want:

xyplot(log.conc ~ time| group, data = foo,
	groups = subject,
	panel = function(..., type, col, lwd, span)
	{
	    panel.grid(h=-1, v=-1, lwd = 1)
	    panel.superpose(..., type = 'l', col = "black", lwd = 1)
	    panel.loess(..., span = .66, col = "red", lwd = 3)
	})

Deepayan