stat_bin accepts functions for binwidth #1890

botanize · 2016-11-03T20:25:26Z

Allowing binwidth to accept functions greatly improves histograms of faceted variables. Furthermore, the current default (30 bins) is somewhat arbitrary, and could be replaced with one of the three built-ins included here.

The built-ins include 'Sturgis', 'Scott', and 'FD'. These match the options included in d3. More information is available on wikipedia.

Examples are included in the pull-request, here's a motivating example:

mtlong <- reshape2::melt(mtcars)
ggplot(mtlong, aes(value)) + geom_histogram() + facet_wrap(~variable, scales = 'free_x')
ggplot(mtlong, aes(value)) + geom_histogram(binwidth = 'Sturges') + facet_wrap(~variable, scales = 'free_x')

Allows variable binwidths across facets.

Sturgis to Sturges

hadley · 2016-11-04T00:53:16Z

I wouldn't mind extending bin to accept functions but I really dislike Sturgis/Scott/FD — I think they give the illusion that you're picking a good binwidth, but they're just as arbitrary as picking 30 bins.

botanize · 2016-11-04T02:17:25Z

None of the options are a default, they have to be chosen manually, which is still a level of intention beyond 30 bins.

Would you prefer to drop the built-in options entirely?

hadley · 2016-11-04T03:09:04Z

Yes please

Better visualize varying ranges.

botanize · 2016-12-14T15:55:04Z

Is there anything else preventing you from accepting this pull request?

hadley

A few more suggestions

hadley · 2017-01-24T23:16:49Z

R/stat-bin.r

@@ -111,9 +113,10 @@ StatBin <- ggproto("StatBin", Stat,
      stop("Only one of `boundary` and `center` may be specified.", call. = FALSE)
    }

-    if (is.null(params$breaks) && is.null(params$binwidth) && is.null(params$bins)) {
+    if (is.null(params$breaks) && (is.null(params$binwidth) || !(is.numeric(params$binwidth) || is.function(params$binwidth))) && is.null(params$bins)) {


Why does this need to be changed? If a function is supplied binwidth will no longer be NULL

Thanks, that was leftover from checking for a string input as well.

hadley · 2017-01-24T23:45:36Z

R/stat-bin.r

@@ -130,8 +133,11 @@ StatBin <- ggproto("StatBin", Stat,
    if (!is.null(breaks)) {
      bins <- bin_breaks(breaks, closed)
    } else if (!is.null(binwidth)) {
-      bins <- bin_breaks_width(scales$x$dimension(), binwidth, center = center,
-        boundary = boundary, closed = closed)
+      if (!is.numeric(binwidth)) {


It would be better to check for is.function() here

hadley · 2017-01-24T23:45:50Z

R/stat-bin.r

-      bins <- bin_breaks_width(scales$x$dimension(), binwidth, center = center,
-        boundary = boundary, closed = closed)
+      if (!is.numeric(binwidth)) {
+        binwidth <- do.call(binwidth, list(data$x))


I don't think you need do.call here

hadley · 2017-01-24T23:46:32Z

tests/testthat/test-stat-bin.R

@@ -22,6 +22,14 @@ test_that("bins specifies the number of bins", {
  expect_equal(nrow(out(bins = 100)), 100)
 })

+test_that("binwidth computes widths for function input", {
+  df <- data.frame(x = 1:100)
+  out <- function(x, ...) {


I think you can remove the indirection of this function now.

And a simpler binwidth function would just return a constant.

Ok, I get the simpler function that just returns a constant, but I don't know what "indirection of this function" means.

Sorry, that was a bit unclear. I meant eliminate the out function:

out <- layer_data(ggplot(...))

hadley · 2017-01-26T17:21:13Z

NEWS.md

@@ -16,6 +16,10 @@

 ## Major new features

+### Binning


Can you please move this up a section?

hadley · 2017-01-26T17:22:16Z

tests/testthat/test-stat-bin.R

@@ -22,6 +22,14 @@ test_that("bins specifies the number of bins", {
  expect_equal(nrow(out(bins = 100)), 100)
 })

+test_that("binwidth computes widths for function input", {
+  df <- data.frame(x = 1:100)
+  out <- function(x, ...) {


Sorry, that was a bit unclear. I meant eliminate the out function:

out <- layer_data(ggplot(...))

hadley · 2017-01-26T19:08:48Z

Thanks!

lock · 2019-01-18T13:20:02Z

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/

Joey Reid added 6 commits November 3, 2016 13:03

Add binwidth functions

aa2b4e6

Allows variable binwidths across facets.

Add binwidth functions to NEWS

3e26bf6

Add examples for binwidth functions

d0b906f

Update documentation

7dafd4d

Add tests for binwidth functions

95fe3b6

Fix type in function name

a4af551

Sturgis to Sturges

Joey Reid added 2 commits November 3, 2016 22:14

Remove predefined binning functions

40d5f46

Set scales to free_x in example

880982c

Better visualize varying ranges.

hadley reviewed Jan 24, 2017

View reviewed changes

Joey Reid added 4 commits January 26, 2017 08:20

Simplify binwidth function test

72be48d

Remove unnecessary checks

6afc2ad

Call function directly

4a25571

Merge remote-tracking branch 'upstream/master' into binwidth-fun

a5e84ff

hadley reviewed Jan 26, 2017

View reviewed changes

Joey Reid added 2 commits January 26, 2017 09:34

Remove function

333f067

Move up a section in NEWS

ca5d16a

hadley merged commit 4c07082 into tidyverse:master Jan 26, 2017

lock bot locked and limited conversation to collaborators Jan 18, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

stat_bin accepts functions for binwidth #1890

stat_bin accepts functions for binwidth #1890

botanize commented Nov 3, 2016

hadley commented Nov 4, 2016

botanize commented Nov 4, 2016

hadley commented Nov 4, 2016

botanize commented Dec 14, 2016

hadley left a comment

hadley Jan 24, 2017

botanize Jan 26, 2017

hadley Jan 24, 2017

botanize Jan 26, 2017

hadley Jan 24, 2017

hadley Jan 24, 2017

botanize Jan 26, 2017

hadley Jan 26, 2017

hadley Jan 26, 2017

hadley Jan 26, 2017

hadley commented Jan 26, 2017

lock bot commented Jan 18, 2019

stat_bin accepts functions for binwidth #1890

stat_bin accepts functions for binwidth #1890

Conversation

botanize commented Nov 3, 2016

hadley commented Nov 4, 2016

botanize commented Nov 4, 2016

hadley commented Nov 4, 2016

botanize commented Dec 14, 2016

hadley left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hadley commented Jan 26, 2017

lock bot commented Jan 18, 2019