how to group smooth line by two groups?

[Lydia Jia]

Hello, I have the following sample dataset.

sex <- as.factor(c(rep(0,12),rep(1,12)))

char <- c("x","x","x","x","y","y","y","y","z","z","z","z","x","x","x","x","y","y","y","y","z","z","z","z")

a <- runif(24,0,2)

b <- runif(24,4,10)

data <- data.frame(cbind(sex,char,a,b))

I would like to generate a blak and white plot with two groups of three smooth lines. linetypes are recognized by sex, and each line with dots, squares and triangles according to the group of char. the following is one group, i need to plot two, one type is solid, the other type is dotted. how to do it?

I tried a color plot with the following code. what should i change "color" to? 

pb <- ggplot(data, aes(x = a, y = b, 

                       color = char, linetype = factor(sex))) 

p0 <- pb +

  scale_color_manual(values = c("#00FFFF", "#FFFF00", "#00FF00")) +

  scale_linetype_manual(breaks=c("0","1"), values=c(1,2), labels = c("male", "female")) +

  geom_smooth(method = "lm", 

              se = FALSE, # Don't add shaded confidence region

              size = 1.2, aes(group = char : sex))

Thanks a looooot!

[Brandon Hurr]

Something like this?

ggplot(data, aes(x = a, y = b, linetype = sex, color=char, group = sex:char))+ 

scale_color_grey() +

scale_linetype_manual(breaks=c("male","female"), values=c(1,2), labels = c("male", "female")) +

geom_line()+

geom_point()+

theme_bw()

--
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

[Lydia Jia]

this is nice! but what i need is regression lines instead of lines connecting scatter points. what should i do?

[Lydia Jia]

and how should i change the shape of the points instead of the color of them?

On Monday, October 22, 2012 10:45:47 AM UTC-4, Brandon Hurr wrote:

[Brandon Hurr]

I was going on the picture and not your code so I didn't use geom_smooth(). The idea is similar though... 

ggplot(data, aes(x = a, y = b, color = char, linetype = factor(sex), shape=char, group=char:sex))+ 

  geom_point(color="black")+

  scale_color_grey() +

  scale_linetype_manual(breaks=c("male","female"), values=c(1,2), labels = c("male", "female")) +

  geom_smooth(method = "lm", 

              se = FALSE, # Don't add shaded confidence region

              size = 1.2)+

  theme_bw()

[Lydia Jia]

sorry i think i didn't make my question clear. 

the points should be on regression lines, just for identity. this means there should be six different line types. but there is no lines type with points that can be shaped. so i need to add points to them.

is there any commands that enable me to add points (shaped by char) onto the regression lines if i specify x value. thanks.

[Brandon Hurr]

Not that I know of. You'll have to use lm() and predict() to make your points on the linear model output of geom_smooth(). 

Is the reason that you're doing that just for identification? Isn't that what the color x linetype tells you? 

[Lydia]

Yes, just for identification. the three colors are great, but i have to identify lines by shape.......

[Lydia]

could you please explain how to use predict() to make the points on the regression line? thanks.

On Monday, October 22, 2012 1:23:52 PM UTC-4, Brandon Hurr wrote:

[Ben Bond-Lamberty]

m <- lm( y ~ x, data=d )

d$predictions <- predict( m )

Ben

[Lydia]

Thanks a lot for answering my question. i will go with scale_color_grey()

but the color of the linetype(solid - male, dotted - female) in the legend is blue.. i think it is default. 

how can i make it black?

thanks!

On Monday, October 22, 2012 1:23:52 PM UTC-4, Brandon Hurr wrote:

[Ben Bond-Lamberty]

It should be simply +geom_line(..., color="black" ) or +geom_smooth( ..., color="black" ) ?

Ben

[Lydia]

but the color is defined by scale_color_manual or scale_color_grey 

so there is a problem: if i add color = black to geom_smooth, it would only give out black lines, not grey and light grey lines. 

[Ben Bond-Lamberty]

Can you provide the exact code you're trying to run? I know you did originally, but as things have changed...

Thanks,

Ben

[Lydia]

    pbaseline <- ggplot(data, aes(x = a, y = b, 

                                            color = char, linetype = sex))

    

    p0 <- pbaseline + 

      scale_color_manual(values = c("#000000", "#666666", "#CCCCCC")) +

      scale_linetype_manual(breaks=c("0","1"), values=c(1,2), labels = c("male", "female")) +

      geom_smooth(method = "lm", se = FALSE, size = 1.2) +  

      labs(linetype = "sex") +

      opts(title = expression("Title", face = "bold", size=20),

           panel.background = theme_rect(fill='white', colour='black'),

           axis.title.x = theme_text(size = 16, colour = 'black'),

           axis.title.y = theme_text(size = 16, colour = 'black', angle = 90),

           legend.text = theme_text(size = 14),

           legend.title = theme_text(face = "bold", size = 16),

           plot.title = theme_text(face = "bold", size = 20)) 

[Brandon Hurr]

Your code wasn't reproducible on my machine and you are using "old" stuff with opts()... The code below works for me, but I do not know how to change the default colors in the legends like you are requesting. I know there is a way as I have seen someone do it before by hacking the legend, but I can't find it in my searches. Hopefully someone who knows can chime in. 

sex <- as.factor(c(rep(0,12),rep(1,12)))

char <- c("x","x","x","x","y","y","y","y","z","z","z","z","x","x","x","x","y","y","y","y","z","z","z","z")

a <- runif(24,0,2)

b <- runif(24,4,10)

data <- data.frame(cbind(sex,char,a,b))

ggplot(data, aes(x = a, y = b, color = char, linetype = sex, group=char:sex))+

      scale_color_manual(values = c("#000000", "#666666", "#CCCCCC")) +

      #scale_linetype_manual(labels = c("male", "female"), guide="legend") +

      geom_smooth(method = "lm", se = FALSE, size = 1.2) +  

      labs(linetype = "sex") + ggtitle(expression("Title", face = "bold", size=20))+

      theme(

           panel.background = element_rect(fill='white', colour='black'),

           axis.title.x = element_text(size = 16, colour = 'black'),

           axis.title.y = element_text(size = 16, colour = 'black', angle = 90),

           legend.text = element_text(size = 14),

           legend.title = element_text(face = "bold", size = 16),

           plot.title = element_text(face = "bold", size = 20))

[Ben Bond-Lamberty]

Simplifying for the moment, this:

pbaseline+geom_smooth(method = "lm", se = FALSE, size =
1.2)+scale_color_manual(values = c("#000000", "#666666", "#CCCCCC"))

produces a plot with the geom_smooth lines black, grey, etc., colored
by 'char'. Is this what you want?

Ben

[Lydia]

this is nice. what i need more is to change the default line color in the "linetype" legend. do you have any idea how to do it? thanks.

[Lydia]

orz.... i didn't suppose it would be so hard... but thanks a lot!!

[Ben Bond-Lamberty]

OK, I didn't expect this to be hard...but don't know, sorry. Maybe
someone more knowledgeable will weigh in.

One workaround would be to get rid of geom_smooth: fit the 'lm'
smoothing lines yourself, building a new data frame based on the model
fits, and then use geom_line to plot it. Presumably geom_line will
respect your color choices better.

Ben

[Charlotte Wickham]

You want to change the aesthetics but only in the guide,

pbaseline +
geom_smooth(method = "lm", se = FALSE, size =1.2) +

scale_color_manual(values = c("#000000", "#666666", "#CCCCCC")) +

scale_linetype(guide = guide_legend(override.aes = list(colour = "black")))


Charlotte

[Lydia]

Yeah, it works!!! Thank you soooooo much!!!!