[R] Achieve independent fine user control of ggplot geom settings when using groups in multiple geom's

Fri Oct 30 00:31:53 CET 2015

Thank for your reply,

I may accept your point about the mapping consistency when the different 
geom's use the same data source. However, as pointed out in my example 
code, this does not have to be the case. Hence my question about the 
geom-specific control of group-dependent graphical settings.

Sebastien

On 10/29/2015 4:49 PM, Jeff Newmiller wrote:
> I think a fundamental design principle of ggplot is that mapping of values to visual representation are consistent within a single plot, so reassigning color mapping for different elements would not be supported.
>
> That being said, it is possible to explicitly control specific attributes within a single geom outside of the mapping, though this usually does break mappings in the legend.
> ---------------------------------------------------------------------------
> Jeff Newmiller                        The     .....       .....  Go Live...
> DCN:<jdnewmil at dcn.davis.ca.us>        Basics: ##.#.       ##.#.  Live Go...
>                                        Live:   OO#.. Dead: OO#..  Playing
> Research Engineer (Solar/Batteries            O.O#.       #.O#.  with
> /Software/Embedded Controllers)               .OO#.       .OO#.  rocks...1k
> ---------------------------------------------------------------------------
> Sent from my phone. Please excuse my brevity.
>
> On October 29, 2015 11:27:55 AM MST, sbihorel <Sebastien.Bihorel at cognigencorp.com> wrote:
>> Thank you for your reply.
>>
>> I do not have anything specific data/geom/grouping in mind, rather a
>> framework in which users would just pile of each other layer after
>> layer
>> of geom each defined with specific settings. A minimum realistic
>> scenario would a geom_point followed by a geom_smooth or a geom_path
>> using different colors...
>>
>> Sebastien
>>
>> On 10/29/2015 1:34 PM, Ista Zahn wrote:
>>> I would say in a word, 'no'. What you seem to be implying is that you
>>> want multiple color scales, multiple shape scales, etc. As far as I
>>> know there is no support for that in ggplot2.
>>>
>>> Perhaps if you show us what you're actually trying to accomplish
>>> someone can suggest a solution or at least a work-around.
>>>
>>> Best,
>>> Ista
>>>
>>> On Thu, Oct 29, 2015 at 12:26 PM, sbihorel
>>> <Sebastien.Bihorel at cognigencorp.com> wrote:
>>>> Hello,
>>>>
>>>> Before I get to my question, I want to make clear that the topic of
>> my
>>>> present post is similar to posts I recently submitted to the list.
>> Although
>>>> I appreciate the replies I got, I believe that I did not correctly
>> frame
>>>> these previous posts to get to the bottom of things.
>>>> I also want to make clear that the code example that I have inserted
>> in this
>>>> post is meant to illustrate my points/questions and does not reflect
>> a
>>>> particular interest in the data or the sequence of ggplot geom's
>> used
>>>> (except otherwise mentioned). Actually, I purposefully used junk
>> meaningless
>>>> data, geom's sequence, and settings, so that we agree the plot is
>> ugly and
>>>> that we, hopefully, don't get hang on specifics and start discussing
>> about
>>>> the merit of one approach vs another.
>>>>
>>>> So here are my questions:
>>>>
>>>> 1- Can a user independently control the settings of each geom's used
>> in a
>>>> ggplot call sequence when grouping is required?
>>>>
>>>> By control, I mean: user defines the graphical settings (groups,
>> symbol
>>>> shapes, colors, fill colors, line types, size scales, and alpha) and
>> does
>>>> not let ggplot choose these settings from some theme default.
>>>> By independently, I mean: the set of graphical settings can be
>> totally
>>>> different from one group to the next and from one geom to the next.
>>>>
>>>> If this fine control can be achieved, how would you go about it
>> (please, be
>>>> assured that I already spent hours miserably failing to get to
>> anything
>>>> remotely productive, so your help would be really appreciated)?
>>>>
>>>> library(dplyr)
>>>> library(tidyr)
>>>> library(ggplot2)
>>>> set.seed(1234)
>>>> dummy <- data.frame(dummy = numeric())
>>>> data <- data.frame(x1 = rep(-2:2, each = 80) + rnorm(4000, sd =
>> 0.1),
>>>>                      g1 = rep(1:4, each = 1000))
>>>> data <- data %>% mutate(y1 = -x1^2 + 2*x1 - 2 + g1 + rnorm(4000, sd
>> = 0.25))
>>>> data2 <- data %>% select(x2=x1, y2=y1, g2=g1) %>% mutate(x2=-x2)
>>>> data3 <- data.frame(x3 = sample(seq(-2, 2, by = 0.1), 20, replace =
>> TRUE),
>>>>                       y3 = runif(20, min=-8, max=4),
>>>>                       g3 = rep(1:4, each = 5)) %>% group_by(g3) %>%
>>>> arrange(x3)
>>>>
>>>> gplot <- ggplot(dummy) ### I know this line is not necessary in this
>>>> particular example, please assume this is relevantin the actual
>> framework I
>>>> am trying to build
>>>> gplot <- gplot +
>>>>     geom_smooth(data = data2,
>>>>                 aes(x2, y2, group = g2, color = factor(g2), linetype
>> =
>>>> factor(g2), size = 0.5*g2),
>>>>                 method = 'loess') +
>>>>     geom_path(data = data3,
>>>>               aes(x3, y3, group = g3, color = factor(g3), linetype =
>>>> factor(g3), shape = factor(g3), size = 0.5*g3)) +
>>>>     geom_point(data = data,
>>>>                aes(x1, y1, group = g1, color = factor(g1), fill =
>> factor(g1),
>>>> shape = factor(g1), size = g1))
>>>> gplot
>>>>
>>>> 2- Is the situation easier or more complex (ie, does ggplot make
>> some
>>>> decisions/assumptions for the user?) if the same x, y, and group
>> variables
>>>> are used in different geom's but the user still wants to provide
>>>> independently graphical settings for each geom?
>>>>
>>>> Thank you
>>>>
>>>> Sebastien
>>>>
>>>> ______________________________________________
>>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>> PLEASE do read the posting guide
>> http://www.R-project.org/posting-guide.html
>>>> and provide commented, minimal, self-contained, reproducible code.
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
>> http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.

-- 
Sebastien Bihorel
Cognigen Corporation
(t) +1 716 633 3463 ext 323
Cognigen Corporation, a wholly owned subsidiary of Simulations Plus, Inc.