[R] plot() and lines() multivariate problem

[Uwe Ligges]

Fred Jean wrote:
> 
> Dear list
> 
> I'm trying to solve the following problem since 2 days with no success.
> Could someone help a newbie, please ?
> 
> I have a  dependant variable which is diameter of shell, and two factors
> which are level on the beach (3 levels 'B', 'E' and 'H') and a hydrodynamics
> indice (3 levels 1, 2 and 3)
> 
> To study diameter variations according to these factors, I may use something
> like :
> > mod <- lm(diameter  ~ level * hydro)
> 
> When I want to look at the graphical results of such a model on a
> scatterplot, I use
> 
> > plot(diameter ~ as.integer(hydro), col = as.integer(level))
> 
> and to draw the model lines :
> 
> > lines(as.integer(hydro)[lev=='B'], predict(mod)[lev=='B'], col = 1)
> > lines(as.integer(hydro)[lev=='E'], predict(mod)[lev=='E'], col = 1)
> > lines(as.integer(hydro)[lev=='H'], predict(mod)[lev=='H'], col = 1)
> 
> But sometimes, the lines are not drawn in the order of the levels of hydro but
> as a *zigzag* and lines() commands seem to draw lines going from level 1
> to level 3 and then back to level 2 of the x axis (hydro)

All data must be sorted according to the hydro variable.


> I hope this description is comprehensible. Could someone explain me how to
> avoid this ?
> 
> Many thanks for giving me some of your time.

You are looking for interaction.plot(), I think, see ?interaction.plot
for details.

Uwe Ligges