NA value in the legend of the plot ?

[Michel ARNAUD]

Hi

I have the following problem :

With these instructions, why NA is it not shown in the legend of the plot ?

qplot(sleep_total, sleep_cycle, data=msleep, colour=vore) + 

scale_colour_hue(

"que mange\nt il ?", 

breaks = c("herbi", "carni", "omni", "insecti", NA),

labels = c("plante", "viande", "tout", "insecte", "Ne sais pas"))

Thanks for your help

Michel

[Brandon Hurr]

If you put NA in "" does it put it in? 

breaks = c("herbi", "carni", "omni", "insecti", "NA"),

B

--
--
You received this message because you are subscribed to the ggplot2 mailing list.
Please provide a reproducible example: https://github.com/hadley/devtools/wiki/Reproducibility
 
To post: email ggp...@googlegroups.com
To unsubscribe: email ggplot2+u...@googlegroups.com
More options: http://groups.google.com/group/ggplot2

---
You received this message because you are subscribed to the Google Groups "ggplot2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ggplot2+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Michel ARNAUD]

same result !!!

Michel

 

Michel Arnaud

5 rue du Cannau

34000 Montpellier

[Brandon Hurr]

I didn't realize that msleep was a dataset in R so I didn't try it, but now that I did I can see the problem. 

Any place where there is NA in the "vore" column, there is missing data in sleep_total or sleep_cycle. When I run the code I get... 

Warning message:

Removed 51 rows containing missing values (geom_point). 

The rows that are being removed from this dataset include the rows that are NA in "vore", because they are NA in one or both of the other columns that you are using geom_point to plot. 

Make sense? 

B

[Michel ARNAUD]

No because

msleep[!is.na(msleep$sleep_total) & !is.na(msleep$ sleep_cycle) ,c(3,6,8)]

      vore sleep_total sleep_cycle

4     omni        14.9   0.1333333

5    herbi         4.0   0.6666667

6    herbi        14.4   0.7666667

7    carni         8.7   0.3833333

9    carni        10.1   0.3333333

12   herbi         9.4   0.2166667

14   herbi        12.5   0.1166667

17    omni         9.1   0.1500000

18   carni        17.4   0.3833333

20    omni        18.0   0.3333333

22 insecti        19.7   0.1166667

23   herbi         2.9   1.0000000

25    omni        10.1   0.2833333

28   carni        12.5   0.4166667

29    omni         9.8   0.5500000

34    omni         8.0   1.5000000

38    omni        10.1   0.7500000

40   herbi        14.3   0.2000000

42   herbi        12.5   0.1833333

43 insecti        19.9   0.2000000

48   herbi         8.4   0.4166667

50    omni         9.7   1.4166667

54    omni         9.4   0.6666667

64   herbi        13.0   0.1833333

67 insecti         8.4   0.1666667

68   herbi        11.3   0.1500000

71   herbi        13.8   0.2166667

73    <NA>        12.8   0.1833333

74    omni         9.1   0.5000000

77   herbi         4.4   0.9000000

79    omni         8.9   0.2333333

83   carni         9.8   0.3500000

Michel Arnaud

5 rue du Cannau

34000 Montpellier

[Brandon Hurr]

Michel, 

I missed that one. It took me a while to realize that it was actually plotting that point in grey, but not putting anything into the legend. 

Best I can do is to convert it to a character and then give it a non-NA name. 

msleep$vore<-as.character(msleep$vore)

msleep[which(is.na(msleep$vore)),]$vore<-"None"

ggplot(data=msleep)+

geom_point(aes(sleep_total, sleep_cycle, colour=vore), size=3)

[Michel ARNAUD]

Merci beaucoup  Brandon

Bonne journée à toi

Michel Arnaud

5 rue du Cannau

34000 Montpellier

[Dennis Murphy]

There is an argument in the scale functions to plot NA values of the
aesthetic in a different color (the default is "gray50" or "grey50",
depending on your preferred spelling); however, it does not print an
entry in the legend because by default, NAs are excluded as legitimate
factor levels. If you use the 'exclude = NULL' argument in factor() to
generate an NA level, it still doesn't get passed to the legend in
ggplot2. Therefore, you need to resort to the 'replace the NA by
something else' in order to have it represented in the legend.

I went about this in a more roundabout way, but got to the same place:

msleep2 <- msleep[!is.na(msleep$sleep_cycle), ]

# Recast vore as a factor with an NA level
msleep2$Vore <- as.character(msleep2$vore)
msleep2$Vore[is.na(msleep2$Vore)] <- "NA"
msleep2$Vore <- factor(msleep2$Vore, levels = c("carni", "herbi", "insecti",
"omni", "NA"))

qplot(sleep_total, sleep_cycle, data=msleep2, colour=Vore) +
scale_colour_manual("que mange\nt il ?",
values = c("firebrick", "brown", "darkorange", "dodgerblue", "black"),

labels = c("plante", "viande", "tout", "insecte", "Ne sais pas"))

Dennis

[Michel ARNAUD]

Merci Denis

Michel