How can I get my points to connect in a plot and show a trend with NA values in data?

问题

I have a basic set of data, where I measure concentration over time for 24 months. Some months have not been sampled, so there are 6 NA values in my list.

The na.omit() function removes my NA values and gives me the graph I am looking for, but it ruins my axis.

(1)

 plot(time, pt, type="o", pch=16, col="blue", 
    xlab="Time (months) relative to implant", ylab="Concentration (ng/ml)", 
    main="Concentration Overtime", 
    xaxt='n')
    axis(1, at=seq(-1, 24, by=1))

(2)

 plot(na.omit(pt), type="o", pch=16, col="blue", xlab="Time (months) relative 
        to implant", ylab="Concentration (ng/ml)", 
        main="Concentration Overtime", 
        xaxt='n')
        axis(1, at=seq(-1, 24, by=1))

My graph looks like this

Plot 1

Plot 2

回答1:

Using data from Gwang-Jin Kim's answer and the suggestion of dashing the lines where NAs exist, here's a redo. I'll keep the previous answer below.

lines (and therefore plot(..., type="l")) requires that lty be the same for all components, so to get dashed sections you need to use segments over each point-to-point.

For extra credit, I'm including red dots along the x-axis where data was dropped at the time the NA value exists.

df <- data.frame(time=c(-1, 0, 1:24),
                 pt=c(7.0, 6.9, NA, 5.5, 5, 3, 14, NA, 23, NA, 14.5, 7, 9, NA,
                      11, 8, 5.2, 5.3, NA, 5, 3, NA, 1.5, NA, NA, 2))
len <- nrow(df)
notna <- !is.na(df$pt)
df$dashes <- c(TRUE, !notna[-len])
df0 <- df[notna,]
len0 <- nrow(df0)

plot(pt ~ time, data=df0,
     type="p", pch=16, col="blue", 
     xlab="Time (months) relative to implant", ylab="Concentration (ng/ml)", 
     main="Concentration Overtime", 
     xaxt='n')
points(df$time, par('usr')[3] * is.na(df$pt), pch = 16, col = "red")
ign <- Map(segments, df0$time[-len0], df0$pt[-len0],
           df0$time[-1], df0$pt[-1],
           1, 1+df0$dashes[-1])
axis(1, at=seq(-1, 24, by=1))

Here's a guess:

notna <- !is.na(pt)
plot(time[notna], pt[notna], type="o", pch=16, col="blue", 
    xlab="Time (months) relative to implant", ylab="Concentration (ng/ml)", 
    main="Concentration Overtime", 
    xaxt='n')
    axis(1, at=seq(-1, 24, by=1))

One problem in your second code

plot(na.omit(pt), ...)

is that you have not included time, so R naturally fills in with seq_along(na.omit(pt)) (almost the same as 1:length(na.omit(pt))), which for you is losing your time data. By using [notna] to subset both vectors, we are preserving the data you need for the plot.

回答2:

df <- data.frame(time=c(-1, 0, 1:24),
pt=c(7.0, 6.9, NA, 5.5, 5, 3, 14, NA, 23, NA, 14.5, 7, 9, NA, 11, 8, 5.2, 5.3, NA, 5, 3, NA, 1.5, NA, NA, 2))

png("test.png")
plot(pt~time, type="o", data=na.omit(data.frame(df)),
     pch=16, col="blue",
     xlab="Time (months) relative to implant",
     ylab="Concentration (ng/ml)",
     main="Concentration Overtime",
     xlim=c(-1, 24),
     xaxt='n')
axis(1, at=seq(-1, 24, by=1))
dev.off()

来源：https://stackoverflow.com/questions/53132402/how-can-i-get-my-points-to-connect-in-a-plot-and-show-a-trend-with-na-values-in

标签

plot

trend