R - Plot interpolated values a different colour in time series plot?

Question

Can someone help me. I have a dataset that had NA values that I have interpolated with zoo. I have added a 'colour column' in the hope that I could create a line plot (time series) with the interpolated values plotted in a different colour to the rest of the line. That is, the segment of the line defined by the point immediately before and immediately after the interpolated point should be red, and not black.

I've attached an example of my table here (where the colour is 'red' defines the values that have been interpolated). I've also put an image of the graph so far and the desired output here too:

https://drive.google.com/folderview?id=0B_eJi0urUAzFM0JBS1ZIbUdGck0&usp=drive_web

This is my code thus far. The 'lines' part of the code is where I hoped to define the colour as the column in the data frame:

par(mfrow=c(2,1), mar=c(4,4.5,2,2), mgp=c(2,0.6,0))
x.limit <- round(range(UN.GRACE.Int$DecimDate), 2)
plot(NULL, type="n", xlim=x.limit, ylim=c(-20, 25), xlab="Year", ylab="GRACE-TWS (cm)",     axes=F)
box(lwd=1.5)
abline(h=0, col="gray50", lty=1)
axis(1, seq(2003, 2012, 1), cex.axis=0.8)
axis(2, seq(-20, 25, 5), las=1, cex.axis=0.8)
minor.tick(nx=4, ny=0, tick.ratio=0.5) 
lines(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1], type="l", lwd=3, col=UN.GRACE.Int[,3])
tws.slope <- round(as.vector(coef(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]))[2]), 2)
tws.sdev <- round(as.vector(coef(summary(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1])))[,  "Std. Error"][2]), 2)
abline(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]), lwd=2.5, lty=2, col=2)
mtext(paste("Trend (cm/year): ", tws.slope, "±", tws.sdev, sep=""), cex=0.8, side=1,  line=-1.1) 

Any help would be appreciated - Thanks

Answer 1

If I understand this correctly, you want the interpolated points to show up with a different color. You can accomplish this using the type="o" option in R, which gives over-plotted lines. Here's some adjusted code that produces the following plot. I took the minor.tick command out because it must have been from a package I don't have, but otherwise it works fine (using R 2.15.3 on my local machine).

You'll notice that I just plot the item directly, rather than calling plot to NULL and then adding in lines. This simplifies the code substantially. You can play with the pch parameter in the plot call to change the symbols used, and also alter the lwd parameters as needed. In fact, you could easily give a different value to pch for the interpolated values, like you did color - it accepts a vector as an argument.

par(mar=c(4,4.5,2,2), mgp=c(2,0.6,0))
x.limit <- round(range(UN.GRACE.Int$DecimDate), 2)
plot(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1],
     type="o",
     pch=18,
     col=UN.GRACE.Int[,3],
     xlim=x.limit,
     ylim=c(-20, 25),
     xlab="Year",
     ylab="GRACE-TWS (cm)",
     axes=F)
box(lwd=1.5)
abline(h=0, col="gray50", lty=1)
axis(1, seq(2003, 2012, 1), cex.axis=0.8)
axis(2, seq(-20, 25, 5), las=1, cex.axis=0.8)
tws.slope <- round(as.vector(coef(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]))[2]), 2)
tws.sdev <- round(as.vector(coef(summary(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1])))[,  "Std. Error"][1]), 2)
abline(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]), lwd=2.5, lty=2, col=2)
mtext(paste("Trend (cm/year): ", tws.slope, "±", tws.sdev, sep=""), cex=0.8, side=1,  line=-1.1)

You could also add the points later if you JUST want to see the points where the data was interpolated. This could be done as follows:

par(mar=c(4,4.5,2,2), mgp=c(2,0.6,0))
x.limit <- round(range(UN.GRACE.Int$DecimDate), 2)
plot(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1],
     type="l",
     pch=18,
     col="black",
     xlim=x.limit,
     ylim=c(-20, 25),
     xlab="Year",
     ylab="GRACE-TWS (cm)",
     axes=F)
box(lwd=1.5)
abline(h=0, col="gray50", lty=1)
axis(1, seq(2003, 2012, 1), cex.axis=0.8)
axis(2, seq(-20, 25, 5), las=1, cex.axis=0.8)
tws.slope <- round(as.vector(coef(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]))[2]), 2)
tws.sdev <- round(as.vector(coef(summary(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1])))[,  "Std. Error"][3]), 2)
abline(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]), lwd=2.5, lty=2, col=2)
mtext(paste("Trend (cm/year): ", tws.slope, "±", tws.sdev, sep=""), cex=0.8, side=1,  line=-1.1)
points(x=UN.GRACE.Int[UN.GRACE.Int$Col.CSR=="red",1],
       y=UN.GRACE.Int[UN.GRACE.Int$Col.CSR=="red",2],
       pch=16,
       col="red")

EDITED TO ADD: This is a way to color the line segments themselves by overplotting the original plot, assuming the distance to be colored is always of length one. It uses a quick'n'dirty for() loop, but it could be made into a function if you wanted.

par(mar=c(4,4.5,2,2), mgp=c(2,0.6,0))
x.limit <- round(range(UN.GRACE.Int$DecimDate), 2)
plot(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1],
     type="l",
     pch=18,
     col="black",
     xlim=x.limit,
     ylim=c(-20, 25),
     xlab="Year",
     ylab="GRACE-TWS (cm)",
     axes=F)
box(lwd=1.5)
abline(h=0, col="gray50", lty=1)
axis(1, seq(2003, 2012, 1), cex.axis=0.8)
axis(2, seq(-20, 25, 5), las=1, cex.axis=0.8)
tws.slope <- round(as.vector(coef(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]))[2]), 2)
tws.sdev <- round(as.vector(coef(summary(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1])))[,  "Std. Error"][5]), 2)
abline(lm(UN.GRACE.Int[,2] ~ UN.GRACE.Int[,1]), lwd=2.5, lty=2, col=2)
mtext(paste("Trend (cm/year): ", tws.slope, "±", tws.sdev, sep=""), cex=0.8, side=1,  line=-1.1)

line_segs <- cbind(lstart=UN.GRACE.Int[which(UN.GRACE.Int$Col.CSR=="red")-1,c("DecimDate","CSR")],
                   lend=UN.GRACE.Int[which(UN.GRACE.Int$Col.CSR=="red")+1,c("DecimDate","CSR")])

for(x in 1:nrow(line_segs)) {

    lines(x=c(line_segs[x,1],line_segs[x,3]),
          y=c(line_segs[x,2],line_segs[x,4]),
          lwd=3,
          col="red")
}