Errorbars/Ribbons

[tbreloff]

I am planning to add a new keyword argument: :errorbars. Some notes:

• I'll probably alias err, error, errorbar.
• I want these to be closely linked to "ribbons", where there is a filled area surrounding the line. I need to decide if ribbon should be its own keyword arg, or if I can somehow use errorbars for both.
• I also need to decide if fillrange overlaps here. Should one pass the numerical values into fillrange, then errorbars could be the "style", for example: plot(y, fillrange=std(y), errorbars = :ribbon)

If anyone has good ideas and/or opinions on how they'd like the interface to look, please post here.

[diegozea]

It would be great to have this for both axis (x and y) and supporting symmetric and asymmetric errors.

[tbreloff]

would be great to have this for both axis

So I don't forget... it would be good to use the existing orientation argument for this purpose, which is currently only used for Plotly histograms.

[tbreloff]

See #178 and https://github.com/tbreloff/ExamplePlots.jl/blob/master/notebooks/errorbars.ipynb

[diegozea]

@tbreloff Is it possible to plot error for y and x in the same plot? i. e.

[tbreloff]

Yes it's easy, but need to decide on the right API. Could pass a tuple, or have a separate xerror, or something else? Type out the command you'd like to write to make this happen ;)

[diegozea]

I was looking at the code... At the moment only symmetric errors are supported, aren't they? Maybe tuples can be used for asymmetric errors (lower and upper limits), and add xerror (since error is yerror).

[tbreloff]

Yeah that's a good idea. Alias error to yerror and add xerror. Then I could remove the need for orientation, and keep access to asymmetric errors. I'll do this next.

On Apr 8, 2016, at 11:01 AM, Diego Javier Zea notifications@github.com wrote:

I was looking at the code... At the moment only symmetric errors are supported, aren't they? Maybe tuples can be used for asymmetric errors (lower and upper limits), and add xerror (since error is yerror).

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub

[tbreloff]

Thanks for the API suggestions. I haven't implemented asymmetric errors yet but it should be an easy extension.

using Plots
pyplot(leg=false, size=(500,300))
n = 10
x = [(rand()+1) * randn(n) + 2i for i in 1:5]
y = [(rand()+1) * randn(n) + i for i in 1:5]

f(v) = 1.96std(v) / sqrt(n)
xerr = map(f, x)
yerr = map(f, y)
x = map(mean, x)
y = map(mean, y)

plot(x, y, xerr = xerr, yerr = yerr, marker = stroke(2, :orange))

[tbreloff]

All too easy ;)

plot(x, y, xerr = (0.5xerr,2xerr), yerr = (0.5yerr,2yerr), marker = stroke(2, :orange))

[diegozea]

Awesome! You're welcome :D

[AsimHDar]

In case someone is looking here for single scatter points and asymmetric error bars (inputs to xerr and yerr need to be arrays):

x = 10
y = 4
xe = (2,3) 
ye=  (2,1)

scatter((x,y), yerr=([ye[1]],ye[2]), xerr=([xe[1]],xe[2]) )

[jabruniessner]

Hi there!

After reading all this, I still do not understand. Is it possible to represent the xerror as a ribbon as well somehow?

I would really like to do this!