ggplot grouping by a dichotomous variable not working

[John T]

I'm fairly new to ggplot2, so this is probably a simple mistake. When I do this:

F3.plot <- ggplot(F3, aes(factor(year), mean(cd), color = factor(failure))) +

        geom_line() +

        geom_point(size=2)

F3.plot

I get the attached graph. I'm trying to graph 2 variables by a third, a dichotomous 0 and 1. I suspect the problem is related to the data types, but can't seem to figure out, so in the interest of time, I'd thought I'd ask. I searched the postings and couldn't find anything. Does anyone know why the data is not separating the third variable, called "failure". I also posting the data types. I know for a fact that the 0 and 1 differ by a considerable amount, so it's not mapping the data. I'm taking the mean of the cd variable at ten time points (years). The dataset is F3. Thanks.

[Brandon Hurr]

I postulate that it is in fact there, but they are overplotting. 

I believe your problem is "mean(cd)", which is taking the mean of the entire column and using that as the y value for all points. 

Could you post a full example dataset so we can check for sure? 

--
--
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.

[Vivek Patil]

As of now, your mean(cd) is probably computing a single value for the variable.... I assume you want to compute means of cd across the two levels of failure. You could either compute the means outside of the ggplot environment or use stat_summary to do it within the call. See http://docs.ggplot2.org/current/stat_summary.html

For example, 

year=c(1:10)

cd=rnorm(10,100,100)

failure=rep(c(0,1),5)

df=data.frame(year,cd,failure)

library(ggplot2)

library(dplyr)

df%>%group_by(year,as.factor(failure))%>%

  summarize(mean=mean(cd))%>%

  ggplot(aes(x=year,y=mean,color=as.factor(failure),group=as.factor(failure)))+geom_line()+geom_point(size=2)

As Brandon mentions, a reproducible dataset will come in very handy.

On Thu, Oct 1, 2015 at 9:59 AM, John T <jtca...@gmail.com> wrote:

--

[john]

Ok, here's a slice of it.

[Brandon Hurr]

A small change to Vivek's code will do for what you need I think. 

require(readr)

require(dplyr)

require(ggplot2)

df <- read_csv("~/Downloads/f3sub.csv")

df%>%group_by(year, failure)%>%

  summarize(mean=mean(cd))%>%

  ggplot(aes(x=year,y=mean,color=as.factor(failure),group=as.factor(failure)))+geom_line()+geom_point(size=2)

When grouping, you don't need to use as.factor(), in fact it seems to break it. 

Error in UseMethod("depth") : 

  no applicable method for 'depth' applied to an object of class "NULL"

[John T]

I'm getting this:

> df%>%group_by(year, failure)%>%

+ summarize(mean=mean(cd))%>%

+ ggplot(aes(x=year,y=mean,color=failure, group=failure)) +

+ geom_line() +

+ geom_point(size=2)

Error in eval(expr, envir, enclos) : object 'year' not found

Not found. Don't have time to work on this right now. I tried including factor for year and got the same error. This is after installing the packages.

[John T]

Tried this:

Still got the flat line, no separation of the groups by failure.

F3 %>% group_by(year, failure) %>% summarize(mean=mean(cd))%>%

        ggplot(., aes(x=year,y=mean, color=as.factor(failure), group=as.factor(failure))) + 

        geom_line()+

        geom_point(size=2)

They are different: 

> sub.df0 <- subset(sub.df$cd, failure == "0")

> mean(sub.df0)

[1] 432.86

> sub.df1 <- subset(sub.df$cd, failure == "1")

> mean(sub.df1)

[1] 466.48

> 

[Vivek Patil]

Instead of ggplot(., aes(.. 

Try ggplot(aes(..

[Vivek Patil]

John, sorry, please disregard my earlier email. Your code works and so should my suggestion.

[John T]

It's not calculating a mean cd by year, just an overall mean, so the plot is just a horizontal line.

[John T]

Two groupings, first mean cd by year, then by a dichotomous variable (0, 1)

On Friday, October 2, 2015 at 12:29:43 AM UTC+7, John T wrote: