Skip to content

Telling plot() the max y value to expect when plotting one distribution and then using lines() to add more distributions

5 messages · FJ M, michael.weylandt at gmail.com (R. Michael Weylandt, R. Michael Weylandt +1 more

Original
FJ M
# 
I am plotting three Pearson Type IV distributions. It looks like I have to plot the distribution with the highest value of y and then use lines() to add the two distributions that are shorter / have lower max values of y. The following code figures out which distribution has the max y value, plots it first and then uses lines for the other two distributions with a series of three if statements. This works. I run R from a batch file that reads the following in a text file.
 
I want to graph dozens of distributions and I am looking for a more elegant way to do this.
 
New to R, experienced C programmer, thanks in advance. Frank



colors <- c("red", "blue", "darkgreen", "gold", "black")
labels <- c("1", "2","3")
pdf("C:\\Users\\Frank\\Documents\\R\\Scripts\\pt4_Graph.pdf")
## load Pearson package
library(PearsonDS)
##range for x axis
no_of_increments<- 100
x <- seq(-0.06, +0.06, length=no_of_increments)
##parameters for the plots of three distributions
mx<- c(1.95, 18.35,1.93)
nux<- c(0.08,-1.02,0.25)
locationx<- c(0.0048,-0.00254,0.00189)
scalex<- c(0.0115,0.082187,0.026675)
## calculate probability density function
hx1<- dpearsonIV(x,m=mx[1],nu=nux[1],location=locationx[1],scale=scalex[1])
hx2<- dpearsonIV(x,m=mx[2],nu=nux[2],location=locationx[2],scale=scalex[2])
hx3<- dpearsonIV(x,m=mx[3],nu=nux[3],location=locationx[3],scale=scalex[3])
##calculate max of each distribtion
maxhx1<- max(hx1)
maxhx2<- max(hx2)
maxhx3<- max(hx3)
maxhx<- max(hx1,hx2,hx3)
maxhx1
maxhx2
maxhx3
maxhx

if (maxhx1==maxhx) 
 {plot(x, hx1 , type="l", lwd=2, col=colors[1], xlab="x value",  ylab="Density", main="pt4")
for (i in 2:3){
  lines(x, dpearsonIV(x,m=mx[i],nu=nux[i],location=locationx[i],scale=scalex[i]), lwd=2, col=colors[i])}
legend("topright", inset=.05, title="Distributions",
  labels, lwd=2, lty=c(1, 1, 1, 1, 2), col=colors)
grid()
}

if (maxhx2==maxhx) {plot(x, hx2 , type="l", lwd=2, xlab="x value",  ylab="Density", main="SPX", col=colors[2])
{
  lines(x, dpearsonIV(x,m=mx[1],nu=nux[1],location=locationx[1],scale=scalex[1]), lwd=2, col=colors[1])
  lines(x, dpearsonIV(x,m=mx[3],nu=nux[3],location=locationx[3],scale=scalex[3]), lwd=2, col=colors[3])
legend("topright", inset=.05, title="Distributions",
  labels, lwd=2, lty=c(1, 1, 1, 1, 2), col=colors)
grid()
}

if (maxhx3==maxhx) 
 {plot(x, hx3 , type="l", lwd=2, col=colors[3], xlab="x value",  ylab="Density", main="SPX")
for (i in 1:2){
  lines(x, dpearsonIV(x,m=mx[i],nu=nux[i],location=locationx[i],scale=scalex[i]), lwd=2, col=colors[i])}
legend("topright", inset=.05, title="Distributions",
  labels, lwd=2, lty=c(1, 1, 1, 1, 2), col=colors)
grid()
}
michael.weylandt at gmail.com (R. Michael Weylandt
# 
Easiest thing to do: use the optional ylim argument to plot (taking values like c(0, maxhx) or possibly a little wider) which will let you set the bounds directly. 

Michael

PS It's possible to get the following to be much more easily extensible using loops and what not, but as its very late in my time zone, I'll wait til the morning to work it out (for fear of making yet another public mistake :-P) For now, note that it's perfectly ok to use

maxhx <- max(hx1, hx2, hx3) 

which will save you a few lines.
On Feb 24, 2012, at 5:58 PM, FJ M <chicagobrownblue at hotmail.com> wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      R. Michael Weylandt
    

    

      
      #
    

  


  

      
I might (re-)format your code as follows -- others will make some
different design decisions. This isn't the most efficient (more
vectorization could be squeezed in there and it's arguably better to
use apply statements but I'd argue you'd loose clarity here) but I
think it shows some useful R idioms that are helpful for someone who
(to use Patrick Burns' phrase) "speaks R with a C accent."
(Incidentally, look up a text called "the R inferno": all sorts of
goodies in there)

library(PearsonDS)

PARAMETERS = data.frame(color = c("red", "blue", "green"), label =
c(1, 2, 3), m = c(1.95, 18.35, 1.93), nu = c(0.08, -1.02, 0.25),
location = c(0.0048,  -0.00254, 0.00189), scale = c(0.0115, 0.082187,
0.026675), stringsAsFactors = FALSE)

x <- seq(-0.06, 0.06, length = 100)

densities <- matrix(NA, nrow = nrow(PARAMETERS), ncol = length(x))
for(i in seq_len(nrow(PARAMETERS))){
  densities[i, ] <- with(PARAMETERS[i, ], dpearsonIV(x, m, nu, location, scale))
}

plot(x, rep(0, length(x)), ylim = c(0, max(densities)), xlab = "x
value", ylab = "Density", main = "SPX", type = "n")
for(i in seq_len(nrow(PARAMETERS))){
  lines(x, densities[i, ], col = PARAMETERS[i, "color"])
}

legend("topright", title = "Distributions", legend = PARAMETERS[,
"label"], lwd = 1, col = PARAMETERS[, "color"])
grid()

Hope this helps,

Michael


On Sat, Feb 25, 2012 at 2:22 AM, R. Michael Weylandt

        
<michael.weylandt at gmail.com> <michael.weylandt at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Peter Ehlers
    

    

      
      #
    

  


  

      
See inline below.

        
On 2012-02-25 12:40, R. Michael Weylandt wrote:

            

      
      
      
    

        
I haven't followed this thread carefully so this may already have been
suggested: can't you just replace the plot(), lines() calls with
something like

  matplot(x,t(densities),type='l',lty=1,col=PARAMETERS$color,
          ylab = "Density", main = "SPX")

?

Peter Ehlers

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      R. Michael Weylandt
    

    

      
      #
    

  


  

      
I had never actually played with matplot before -- thanks for the great tip.

Michael

        
On Sat, Feb 25, 2012 at 7:39 PM, Peter Ehlers <ehlers at ucalgary.ca> wrote: