datapoints shape & color

[IRich]

Hello,

I have this dot-plot:

dotAW<-A_W_point

dotAW<-ggplot(dotAW,aes(x=AASW, y=WW, fill=taxa))  

dotAW<-dotAW  + geom_jitter(aes(color = taxa), size = 4)

dotAW<- dotAW  +geom_abline(intercept = 0, slope = 1)

dotAW<- dotAW + scale_y_continuous(limits=c(0,0.5)) + scale_x_continuous(limits=c(0,0.5))

dotAW

ggsave(dotAW, file="dotAW.jpg", width=12, height=10)

data: http://goo.gl/8dZoAq

it's based on relative abundance of bacteria (x & y) and taxa contains a list of operational taxonomic units. I want to correlate it for different the samples (x & y, continuos values). I have so many different taxa, that it's getting confusing, that's why I want to give subset my taxa and give them different shapes, but still using a defined color vector. How can I do that?

Thanks!

Inga 

[Ben Bond-Lamberty]

Yes, it's confusing because you're assigning color to taxa which has
32 different values! If you want to fit and show different linear
regressions use something like
>geom_smooth( method='lm',ads(group=shape ).

Anyway, it might help the list help you to have a better understanding
of what you're trying to do.
Kind regards,
Ben

> --
> --
> 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.

[IRich]

I don't want to do any statistical regression. What I want to visualize is which of the taxa is more abundant when doing a pairwise comparison of my groups. The abline helps for visualization

I hope that clarifies a bit

regards

Inga

[Ben Bond-Lamberty]

If you're committed to an x-y plot, perhaps treat the taxa as a
numeric variable, e.g.
> p <- ggplot(df,aes(x=AASW, y=WW))
> p <- p + geom_jitter(aes(color = as.numeric(taxa), shape=shape), size = 4)
> p <- p +geom_abline(intercept = 0, slope = 1)
> p <- p+ scale_y_continuous(limits=c(0,0.5)) + scale_x_continuous(limits=c(0,0.5))

It all depends on what you're trying to distinguish I guess. Could
also label the points using geom_text and plot everything on a log-log
scale?

Ben

[Dennis Murphy]

It's not obvious to me what you want, but if you want to group your
taxa and identify them with different shapes, here's one way you might
do it, which puts all of the groups on one plot.

ggplot(dotAW, aes(x = AASW, y = WW)) +
geom_point(aes(shape = shape),
position = position_jitter(width = 0.005, height = 0.005)) +
geom_abline(intercept = 0, slope = 1) +
scale_shape_manual(values = c(1, 0, 2),
labels = c("Group 1", "Group 2", "Group 3")) +
labs(shape = "Taxa\ngroups")

I had no idea how you wanted to name the subgroups, so I just threw
something out there; edit it to something sensible.

Another option is to facet by taxa subgroups. If you want to do this,
I'd change the values in dotAW$shape to something more relevant as the
values become the strip titles in the faceted plot.

ggplot(dotAW, aes(x = AASW, y = WW)) +
geom_point(position = position_jitter(width = 0.005, height = 0.005)) +
geom_abline(intercept = 0, slope = 1) + facet_wrap(~ shape, nrow = 1)


Dennis

[IRich]

That looks the way I was thinking. Thank you Dennis!

Do you know there is a way how to additionally color code the groups. I thought of one color vector from color brewer. Maybe it would be to much information in one plot, but I want to try how it looks.

regards

Inga

[IRich]

and thanks Ben as well!

I removed the geom_jitter, it changes the position of the dots in the graphs to avoid overlapping. It is not necessary for my purpose.

/Inga

[Dennis Murphy]

Just add a color aesthetic, but you probably want to merge the two
legends so you have to make sure they have the same breaks, labels and
legend titles.

library(ggplot2)

ggplot(dotAW, aes(x = AASW, y = WW)) +

# theme_bw() +
geom_point(aes(shape = shape, color = shape),

position = position_jitter(width = 0.005, height = 0.005)) +
geom_abline(intercept = 0, slope = 1) +

scale_color_brewer(palette = "Set1",

labels = c("Group 1", "Group 2", "Group 3")) +
scale_shape_manual(values = c(1, 0, 2),
labels = c("Group 1", "Group 2", "Group 3")) +

labs(shape = "Taxa\ngroups", color = "Taxa\ngroups")

This doesn't seem like an improvement to me; to make the colors stand
out, it's probably better to use solid shapes rather than hollow ones,
but the tradeoff is more visual clutter and obscurity in the lower
left corner of the plot.

Dennis

[IRich]

I modified in this following way, doesn't look so bad. It's getting more and more in shape as I had in mind. The overall aim with this color and shape thing is that I would like to have an individual combination of shape and color for each taxa and show this also in the legend. thanks for the patience! I'm a bit lost with combining shape and colors.

/Inga

ggplot(dotAW, aes(x = AASW, y = WW)) + 

     theme_bw() + 

  geom_point(aes(shape = shape, color = shape), size=3,

             position = position_jitter(width = 0.002, height = 0.002))  + xlim(0,0.15) + ylim(0,0.15) +          

  geom_abline(intercept = 0, slope = 1) + 

  scale_color_brewer(palette = "Set1") 

  #labs(shape = "Taxa\ngroups", color = "Taxa\ngroups") 

                     #labels = c("Group 1", "Group 2", "Group 3")) + 

  #scale_shape_manual(values = c(1, 0, 2), 

                     #labels = c("Group 1", "Group 2", "Group 3")) + 

[IRich]

and another thing which is important is that the combination of color and shape remains consistent. I do have different groups which I want to treat the same way. This is also the reason why some of the taxa have zero abundance in this particular plot.

/Inga

[IRich]

can somebody help me with my last question from friday? I get two different legends when I add shape and color.

Thanks so much!
Inga

[Dennis Murphy]

Does this work?

ggplot(dotAW, aes(x = AASW, y = WW)) +
theme_bw() +
geom_point(aes(shape = shape, color = shape), size=3,

position = position_jitter(width = 0.005, height = 0.005)) +

xlim(-0.01, 0.15) + ylim(-0.01, 0.15) +
coord_equal() +

geom_abline(intercept = 0, slope = 1) +
scale_color_brewer(palette = "Set1",
labels = c("Group 1", "Group 2", "Group 3")) +

scale_shape_manual(values = c(16, 15, 17),

labels = c("Group 1", "Group 2", "Group 3")) +
labs(shape = "Taxa\ngroups", color = "Taxa\ngroups")

To reiterate from an earlier message, in order to merge scales, each
one has to have:
* the same legend title;
* the same breaks;
* the same labels.
Since both scales use the same aesthetic, the breaks will be the same
so they don't have to be stated explicitly; however, the labels do
have to be stated in each scale since they differ from their defaults
(the factor levels, which in this case are circle, rectangle and
triangle). Setting the two legend titles in labs() ensures the same
legend titles.

Since you use point jittering, you need to expand your x- and y-limits
on the low end to avoid having jittered points removed from the plot.
With this specification, only the two outliers are removed from the
plot (as opposed to 13 or 14 from your last code chunk).

Dennis

[IRich]

Thanks Dennis! I needed a break from this problem, but I see what you mean and will try to adapt it 

[IRich]

I found a way to go an to merge color and shapes in the legend by using:

guides(colour = guide_legend(override.aes = list(shape = c(16, 16, 16, 16, 16, 16, 16, 16, 16, 15, 15, 15, 15, 15, 15, 15, 15 ,15, 17, 17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18 ,18 ))))

but the way I use it is wrong, it doesn't work

the script:

# made my own color pallete

IRPalette1 <- c("#d73027", "#f46d43", "#fdae61", "#fee090", "#ffffbf", "#e0f3f8", "#abd9e9", "#74add1", "#4575b4")

ggplot(dotAW, aes(x = AASW, y = WW)) + 

  theme_bw() + 

  geom_point(aes(shape = shape, color = co), size=3, 

             position = position_jitter(width = 0.005, height = 0.005))  + 

  xlim(-0.01, 0.15) + ylim(-0.01, 0.15) + 

  coord_equal() + 

  geom_abline(intercept = 0, slope = 1) + 

  scale_color_manual(values =IRPalette2) +                  

  scale_shape_manual(values = c(16, 15, 17, 18)) + 

  guides(colour = guide_legend(override.aes = list(shape = c(16, 16, 16, 16, 16, 16, 16, 16, 16, 15, 15, 15, 15, 15, 15, 15, 15 ,15, 17, 17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18 ,18 ))))

 

I also updated the datafile: https://docs.google.com/spreadsheet/ccc?key=0AnwnLL3Lfn-jdHBQSWg0MUJSS1d4b2RsQlVENlhjRGc&usp=sharing

[IRich]

This finally solved it: http://stackoverflow.com/questions/5435883/how-to-merge-colour-and-shape

/Inga