ggpmisc 0.2.10
Additional examples

Preliminaries

library(ggpmisc)
library(ggplot2)
library(xts)
library(lubridate)
library(tibble)
library(nlme)

Introduction

Many of the functions, including ggplot statistics and geoms, included in package ‘ggpmisc’ had their origin in my need to produce plots for use in teaching. Some of them are more generally useful, such as stat_poly_eq(), but others like stat_fit_deviations() are squarely aimed and producing learning material. Finally, several statistics for debugging and learning how ggplot statistics and geoms interact with each other will be of use only to developers of new statistics and geoms. Function try_data_frame() opens the door to easily converting time series objects into data frames for plotting with ggplot().

try_data_frame()

class(austres)

## [1] "ts"

austres.df <- try_data_frame(austres)
class(austres.df)

## [1] "tbl_df"     "tbl"        "data.frame"

lapply(austres.df, "class")

## $time
## [1] "POSIXct" "POSIXt" 
## 
## $V.austres
## [1] "numeric"

head(austres.df, 4)

## # A tibble: 4 x 2
##         time V.austres
##       <time>     <dbl>
## 1 1971-04-01   13067.3
## 2 1971-07-01   13130.5
## 3 1971-10-01   13198.4
## 4 1972-01-01   13254.2

austres.df <- try_data_frame(austres, as.numeric = TRUE)
lapply(austres.df, "class")

## $time
## [1] "numeric"
## 
## $V.austres
## [1] "numeric"

head(austres.df, 4)

## # A tibble: 4 x 2
##       time V.austres
##      <dbl>     <dbl>
## 1 1971.247   13067.3
## 2 1971.496   13130.5
## 3 1971.748   13198.4
## 4 1972.000   13254.2

class(lynx)

## [1] "ts"

lynx.df <- try_data_frame(lynx)
class(lynx.df)

## [1] "tbl_df"     "tbl"        "data.frame"

lapply(lynx.df, "class")

## $time
## [1] "POSIXct" "POSIXt" 
## 
## $V.lynx
## [1] "numeric"

head(lynx.df, 3)

## # A tibble: 3 x 2
##                  time V.lynx
##                <time>  <dbl>
## 1 1821-01-01 00:00:01    269
## 2 1822-01-01 00:00:01    321
## 3 1823-01-01 00:00:01    585

lynx.df <- try_data_frame(lynx, "year")
head(lynx.df, 3)

## # A tibble: 3 x 2
##         time V.lynx
##       <time>  <dbl>
## 1 1821-01-01    269
## 2 1822-01-01    321
## 3 1823-01-01    585

lynx_n.df <- try_data_frame(lynx, "year", as.numeric = TRUE)
lapply(lynx_n.df, "class")

## $time
## [1] "numeric"
## 
## $V.lynx
## [1] "numeric"

head(lynx_n.df, 3)

## # A tibble: 3 x 2
##    time V.lynx
##   <dbl>  <dbl>
## 1  1821    269
## 2  1822    321
## 3  1823    585

try_data_frame(1:5)

## # A tibble: 5 x 1
##   value
##   <int>
## 1     1
## 2     2
## 3     3
## 4     4
## 5     5

try_data_frame(letters[1:5])

## # A tibble: 5 x 1
##   value
##   <chr>
## 1     a
## 2     b
## 3     c
## 4     d
## 5     e

try_data_frame(factor(letters[1:5]))

## # A tibble: 5 x 1
##    value
##   <fctr>
## 1      a
## 2      b
## 3      c
## 4      d
## 5      e

try_data_frame(list(x = rep(1,5), y = 1:5))

## # A tibble: 5 x 2
##       x     y
##   <dbl> <int>
## 1     1     1
## 2     1     2
## 3     1     3
## 4     1     4
## 5     1     5

try_data_frame(data.frame(x = rep(1,5), y = 1:5))

## # A tibble: 5 x 2
##       x     y
##   <dbl> <int>
## 1     1     1
## 2     1     2
## 3     1     3
## 4     1     4
## 5     1     5

try_data_frame(matrix(1:10, ncol = 2))

## # A tibble: 5 x 2
##      V1    V2
##   <int> <int>
## 1     1     6
## 2     2     7
## 3     3     8
## 4     4     9
## 5     5    10

stat_peaks() and stat_valleys()

Using POSIXct for time and the default formatting of labels.

ggplot(lynx.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "text", colour = "red", vjust = -0.5) +
  ylim(-100, 7300)

Using numeric values for time and the default formatting of labels.

ggplot(lynx_n.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "text", colour = "red", vjust = -0.5) +
  ylim(-100, 7300)

Using POSIXct for time but supplying a format string. In addition marking both peaks and valleys.

ggplot(lynx.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "text", colour = "red", vjust = -0.5, x.label.fmt = "%Y") +
  stat_valleys(colour = "blue") +
  stat_valleys(geom = "text", colour = "blue", vjust = 1.5, x.label.fmt = "%Y") +
  ylim(-100, 7300)

Using numeric for time but supplying a format string. In addition marking both peaks and valleys.

ggplot(lynx_n.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "text", colour = "red", vjust = -0.5, x.label.fmt = "%4.0f") +
  stat_valleys(colour = "blue") +
  stat_valleys(geom = "text", colour = "blue", vjust = 1.5, x.label.fmt = "%4.0f") +
  ylim(-100, 7300)

Rotating the labels.

ggplot(lynx.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "text", colour = "red", angle = 66,
             hjust = -0.1, x.label.fmt = "%Y") +
  ylim(NA, 7800)

Using geom_rug for the peaks and valleys.

ggplot(lynx.df, aes(time, V.lynx)) + geom_line() + 
  stat_peaks(colour = "red") +
  stat_peaks(geom = "rug", colour = "red") +
  stat_valleys(colour = "blue") +
  stat_valleys(geom = "rug", colour = "blue")

stat_poly_eq()

We generate some artificial data.

set.seed(4321)
# generate artificial data
x <- 1:100
y <- (x + x^2 + x^3) + rnorm(length(x), mean = 0, sd = mean(x^3) / 4)
my.data <- data.frame(x, 
                      y, 
                      group = c("A", "B"), 
                      y2 = y * c(0.5,2),
                      block = c("a", "a", "b", "b"))

First one example using defaults.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(formula = formula, parse = TRUE)

stat_poly_eq() makes available three different labels in the returned data frame. One of these is used by default, but aes() can be used to select a different one.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..adj.rr.label..), formula = formula, 
               parse = TRUE)

BIC and AIC labels are also returned.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..AIC.label..), 
               formula = formula, 
               parse = TRUE)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..), formula = formula, 
               parse = TRUE)

Within aes() it is possible to compute new labels based on those returned plus “arbitrary” text. The supplied labels are meant to be parsed into expressions, so any text added should be valid for a string that will be parsed.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label =  paste(..eq.label.., ..adj.rr.label.., sep = "~~~~")),
               formula = formula, parse = TRUE)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = paste("atop(", ..AIC.label.., ",", ..BIC.label.., ")", sep = "")), 
               formula = formula, 
               parse = TRUE)

Two examples of removing and changing the lhs and/or rhs of the equation.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               eq.with.lhs = FALSE,
               formula = formula, parse = TRUE)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               eq.with.lhs = "italic(hat(y))~`=`~",
               formula = formula, parse = TRUE)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  labs(x = expression(italic(z)), y = expression(italic(h)) ) + 
  stat_poly_eq(aes(label = ..eq.label..),
               eq.with.lhs = "italic(h)~`=`~",
               eq.x.rhs = "~italic(z)",
               formula = formula, parse = TRUE)

As any valid R expression can be used, Greek letters are also supported, as well as the inclusion in the label of variable transformations used in the model formula.

formula <- y ~ poly(x, 2, raw = TRUE)
ggplot(my.data, aes(x, log10(y + 1e6))) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               eq.with.lhs = "plain(log)[10](italic(y)+10^6)~`=`~",
               formula = formula, parse = TRUE)

A couple of additional examples of polynomials of different orders, and specified in different ways.

Higher order polynomial.

formula <- y ~ poly(x, 5, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..), formula = formula, parse = TRUE)

Intercept forced to zero.

formula <- y ~ x + I(x^2) + I(x^3) - 1
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..), formula = formula, 
               parse = TRUE)

We give below several examples to demonstrate how other components of the ggplot object affect the behaviour of this statistic.

Facets work as expected either with fixed or free scales. Although bellow we had to adjust the size of the font used for the equation. In addition to we manually position the equation label by supplying coordinates.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..), size = rel(3),
               formula = formula, parse = TRUE) +
  facet_wrap(~group)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..), size = rel(3),
               formula = formula, parse = TRUE) +
  facet_wrap(~group, scales = "free_y")

Grouping, in this example using the colour aesthetic also works as expected. We can use justification and supply an absolute location for the equation.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2, colour = group)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               formula = formula, parse = TRUE) +
  theme_bw()

Label positions relative to the ranges of the x and y scales are also supported, both through string constants and numeric values in the range 0 to 1.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2, colour = group)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               formula = formula, parse = TRUE, label.y.npc = "center") +
  theme_bw()

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2, colour = group)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..eq.label..),
               formula = formula, parse = TRUE, label.y.npc = 0.75) +
  theme_bw()

The default locations are now based on normalized coordinates, and consequently these defaults work even when the range of the x and y scales varies from panel to panel.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2, fill = block)) +
  geom_point(shape = 21, size = rel(3)) +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..rr.label..), size = rel(3),
               geom = "label", alpha = 0.33,
               formula = formula, parse = TRUE) +
  facet_wrap(~group, scales = "free_y") +
  theme_bw()

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y2, colour = group, fill = block)) +
  geom_point(shape = 21, size = rel(3)) +
  geom_smooth(method = "lm", formula = formula) +
  stat_poly_eq(aes(label = ..rr.label..), size = rel(3),
               geom = "label", alpha = 0.2,
               formula = formula, parse = TRUE,
               label.y.npc = 0.66) +
  facet_wrap(~group, scales = "free_y") +
  theme_bw()

stat_fit_residuals

I had the need to quickly plot residuals matching fits plotted with geom_smooth() using grouping and facets, so a new (simple) statistic was born.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = formula)

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = formula, resid.type = "working")

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y, color = group)) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  stat_fit_residuals(formula = formula)

stat_fit_deviations

As I also had the need to highlight residuals in slides and notes to be used in teaching, another statistic was born.

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_deviations(formula = formula, color = "red") +
  geom_point()

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y, color = group)) +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_deviations(formula = formula) +
  geom_point()

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_deviations(formula = formula, color = "red",
                      arrow = arrow(length = unit(0.015, "npc"), 
                                   ends = "both")) +
  geom_point()

stat_fit_glance

# formula <- y ~ poly(x, 3, raw = TRUE)
# broom::augment does not handle poly correctly!
formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_glance(method = "lm", 
                  method.args = list(formula = formula),
                  geom = "text",
                  aes(label = signif(..p.value.., digits = 4)))

# formula <- y ~ poly(x, 3, raw = TRUE)
# broom::augment does not handle poly() correctly!
formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y, color = group)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_glance(method = "lm", 
                  method.args = list(formula = formula),
                  geom = "text", 
                  aes(label = paste("P-value = ", signif(..p.value.., digits = 4), sep = "")))

# formula <- y ~ poly(x, 3, raw = TRUE)
# broom::augment does not handle poly correctly!
formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y, color = group)) +
  geom_point() +
  geom_smooth(method = "lm", formula = formula) +
  stat_fit_glance(method = "lm", 
                  method.args = list(formula = formula),
                  label.x.npc = "right",
                  label.y.npc = "bottom",
                  geom = "text", 
                  aes(label = paste("P-value = ", signif(..p.value.., digits = 4), sep = "")))

stat_fit_augment

Experimental! Use ggplot2::stat_smooth instead of stat_fit_augment if possible.

For a single panel and no grouping, there is little advantage in using this statistic compared to the examples in the documentation of package ‘broom’. With grouping and faceting stat_fit_augment may occasionally be more convenient than ggplot2::stat_smooth because of its flexibility.

# formula <- y ~ poly(x, 3, raw = TRUE)
# broom::augment does not handle poly correctly!
formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  stat_fit_augment(method = "lm",
                   method.args = list(formula = formula))

formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y, color = group)) +
  geom_point() +
  stat_fit_augment(method = "lm", 
                   method.args = list(formula = formula))

We can override the variable returned as to be any of the variables in the data frame returned by broom::augment while still preserving the original y values.

formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y)) +
  stat_fit_augment(method = "lm",
                   method.args = list(formula = formula),
                   geom = "point",
                   y.out = ".resid")

formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y, color = group)) +
  stat_fit_augment(method = "lm",
                   method.args = list(formula = formula),
                   geom = "point",
                   y.out = ".std.resid")

We can use any model fitting method for which augment is implemented.

args <- list(formula = y ~ k * e ^ x,
             start = list(k = 1, e = 2))
ggplot(mtcars, aes(wt, mpg)) +
  geom_point() +
  stat_fit_augment(method = "nls",
                   method.args = args)

args <- list(formula = y ~ k * e ^ x,
             start = list(k = 1, e = 2))
ggplot(mtcars, aes(wt, mpg)) +
  stat_fit_augment(method = "nls",
                   method.args = args,
                   geom = "point",
                   y.out = ".resid")

args <- list(model = y ~ SSlogis(x, Asym, xmid, scal),
             fixed = Asym + xmid + scal ~1,
             random = Asym ~1 | group,
             start = c(Asym = 200, xmid = 725, scal = 350))
ggplot(Orange, aes(age, circumference, color = Tree)) +
  geom_point() +
  stat_fit_augment(method = "nlme",
                   method.args = args,
                   augment.args = list(data = quote(data)))

Debugging ggplots

These stats are very simple and print a summary of their data input to the console. In addition they also return a data frame containing labels suitable for plotting as with geom “text” or geom “label”. However, starting from version 0.2.7 of the package the default geom is “null”. The values are listed to the console at the time when the ggplot object is printed.

As shown here, no other geom or stat is required, however in the remaining examples we include geom_point() to make the data on the plot visible.

ggplot(my.data, aes(x, y)) + stat_debug_group()

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 100 x 4
##        x          y PANEL group
## *  <dbl>      <dbl> <int> <int>
## 1      1 -27205.450     1    -1
## 2      2 -14242.651     1    -1
## 3      3  45790.918     1    -1
## 4      4  53731.420     1    -1
## 5      5  -8028.578     1    -1
## 6      6 102863.943     1    -1
## 7      7 -18547.282     1    -1
## 8      8  13080.521     1    -1
## 9      9  79924.325     1    -1
## 10    10 -44711.499     1    -1
## # ... with 90 more rows

In the absence of facets or groups we get just get the summary from one data frame.

ggplot(my.data, aes(x, y)) + geom_point() + stat_debug_group()

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 100 x 4
##        x          y PANEL group
## *  <dbl>      <dbl> <int> <int>
## 1      1 -27205.450     1    -1
## 2      2 -14242.651     1    -1
## 3      3  45790.918     1    -1
## 4      4  53731.420     1    -1
## 5      5  -8028.578     1    -1
## 6      6 102863.943     1    -1
## 7      7 -18547.282     1    -1
## 8      8  13080.521     1    -1
## 9      9  79924.325     1    -1
## 10    10 -44711.499     1    -1
## # ... with 90 more rows

ggplot(my.data, aes(x, y)) + geom_point() + stat_debug_panel()

## [1] "Input 'data' to 'compute_panel()':"
## # A tibble: 100 x 4
##        x          y PANEL group
##    <dbl>      <dbl> <int> <int>
## 1      1 -27205.450     1    -1
## 2      2 -14242.651     1    -1
## 3      3  45790.918     1    -1
## 4      4  53731.420     1    -1
## 5      5  -8028.578     1    -1
## 6      6 102863.943     1    -1
## 7      7 -18547.282     1    -1
## 8      8  13080.521     1    -1
## 9      9  79924.325     1    -1
## 10    10 -44711.499     1    -1
## # ... with 90 more rows

In the case of grouping then one data frame is summarized for each group in the ggplot object.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_debug_group()

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x          y colour PANEL group
## *  <dbl>      <dbl> <fctr> <int> <int>
## 1      1 -27205.450      A     1     1
## 2      3  45790.918      A     1     1
## 3      5  -8028.578      A     1     1
## 4      7 -18547.282      A     1     1
## 5      9  79924.325      A     1     1
## 6     11  -2823.736      A     1     1
## 7     13 -78016.690      A     1     1
## 8     15 -74281.234      A     1     1
## 9     17   9903.674      A     1     1
## 10    19 -94022.623      A     1     1
## # ... with 40 more rows
## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x         y colour PANEL group
## *  <dbl>     <dbl> <fctr> <int> <int>
## 1      2 -14242.65      B     1     2
## 2      4  53731.42      B     1     2
## 3      6 102863.94      B     1     2
## 4      8  13080.52      B     1     2
## 5     10 -44711.50      B     1     2
## 6     12  23839.55      B     1     2
## 7     14  75601.96      B     1     2
## 8     16 104676.72      B     1     2
## 9     18 -68746.93      B     1     2
## 10    20 -39230.19      B     1     2
## # ... with 40 more rows

Without facets, we still have only one panel.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_debug_panel()

## [1] "Input 'data' to 'compute_panel()':"
## # A tibble: 100 x 5
##        x          y colour PANEL group
##    <dbl>      <dbl> <fctr> <int> <int>
## 1      1 -27205.450      A     1     1
## 2      2 -14242.651      B     1     2
## 3      3  45790.918      A     1     1
## 4      4  53731.420      B     1     2
## 5      5  -8028.578      A     1     1
## 6      6 102863.943      B     1     2
## 7      7 -18547.282      A     1     1
## 8      8  13080.521      B     1     2
## 9      9  79924.325      A     1     1
## 10    10 -44711.499      B     1     2
## # ... with 90 more rows

The data are similar, except for the column named after the aesthetic, for the aesthetics used for grouping.

ggplot(my.data, aes(x, y, shape = group)) + geom_point() + 
  stat_debug_group()

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x          y  shape PANEL group
## *  <dbl>      <dbl> <fctr> <int> <int>
## 1      1 -27205.450      A     1     1
## 2      3  45790.918      A     1     1
## 3      5  -8028.578      A     1     1
## 4      7 -18547.282      A     1     1
## 5      9  79924.325      A     1     1
## 6     11  -2823.736      A     1     1
## 7     13 -78016.690      A     1     1
## 8     15 -74281.234      A     1     1
## 9     17   9903.674      A     1     1
## 10    19 -94022.623      A     1     1
## # ... with 40 more rows
## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x         y  shape PANEL group
## *  <dbl>     <dbl> <fctr> <int> <int>
## 1      2 -14242.65      B     1     2
## 2      4  53731.42      B     1     2
## 3      6 102863.94      B     1     2
## 4      8  13080.52      B     1     2
## 5     10 -44711.50      B     1     2
## 6     12  23839.55      B     1     2
## 7     14  75601.96      B     1     2
## 8     16 104676.72      B     1     2
## 9     18 -68746.93      B     1     2
## 10    20 -39230.19      B     1     2
## # ... with 40 more rows

If we use as geom "label" or "text" a debug summary is added to the plot itself, we can use other arguments valid for the geom used, in this case vjust.

ggplot(my.data, aes(x, y, shape = group)) + geom_point() + 
  stat_debug_group(geom = "label", vjust = c(-0.5,1.5))

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x          y  shape PANEL group
## *  <dbl>      <dbl> <fctr> <int> <int>
## 1      1 -27205.450      A     1     1
## 2      3  45790.918      A     1     1
## 3      5  -8028.578      A     1     1
## 4      7 -18547.282      A     1     1
## 5      9  79924.325      A     1     1
## 6     11  -2823.736      A     1     1
## 7     13 -78016.690      A     1     1
## 8     15 -74281.234      A     1     1
## 9     17   9903.674      A     1     1
## 10    19 -94022.623      A     1     1
## # ... with 40 more rows
## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 50 x 5
##        x         y  shape PANEL group
## *  <dbl>     <dbl> <fctr> <int> <int>
## 1      2 -14242.65      B     1     2
## 2      4  53731.42      B     1     2
## 3      6 102863.94      B     1     2
## 4      8  13080.52      B     1     2
## 5     10 -44711.50      B     1     2
## 6     12  23839.55      B     1     2
## 7     14  75601.96      B     1     2
## 8     16 104676.72      B     1     2
## 9     18 -68746.93      B     1     2
## 10    20 -39230.19      B     1     2
## # ... with 40 more rows

The summary function can be a user defined one, which allows lots of flexibility.

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = summary)

## [1] "Input 'data' to 'compute_group()':"
##        x                y               PANEL       group   
##  Min.   :  1.00   Min.   : -94023   Min.   :1   Min.   :-1  
##  1st Qu.: 25.75   1st Qu.:  40345   1st Qu.:1   1st Qu.:-1  
##  Median : 50.50   Median : 154036   Median :1   Median :-1  
##  Mean   : 50.50   Mean   : 266433   Mean   :1   Mean   :-1  
##  3rd Qu.: 75.25   3rd Qu.: 422069   3rd Qu.:1   3rd Qu.:-1  
##  Max.   :100.00   Max.   :1077469   Max.   :1   Max.   :-1

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = head)

## [1] "Input 'data' to 'compute_group()':"
##   x          y PANEL group
## 1 1 -27205.450     1    -1
## 2 2 -14242.651     1    -1
## 3 3  45790.918     1    -1
## 4 4  53731.420     1    -1
## 5 5  -8028.578     1    -1
## 6 6 102863.943     1    -1

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = nrow)

## [1] "Input 'data' to 'compute_group()':"
## [1] 100

The default.

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = as_data_frame)

## [1] "Input 'data' to 'compute_group()':"
## # A tibble: 100 x 4
##        x          y PANEL group
## *  <dbl>      <dbl> <int> <int>
## 1      1 -27205.450     1    -1
## 2      2 -14242.651     1    -1
## 3      3  45790.918     1    -1
## 4      4  53731.420     1    -1
## 5      5  -8028.578     1    -1
## 6      6 102863.943     1    -1
## 7      7 -18547.282     1    -1
## 8      8  13080.521     1    -1
## 9      9  79924.325     1    -1
## 10    10 -44711.499     1    -1
## # ... with 90 more rows

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = head, summary.fun.args = list(n = 3))

## [1] "Input 'data' to 'compute_group()':"
##   x         y PANEL group
## 1 1 -27205.45     1    -1
## 2 2 -14242.65     1    -1
## 3 3  45790.92     1    -1

This next chunk showing how to print the whole data frame is not run as its output is more than 100 lines long as the data set contains 100 observations.

ggplot(my.data, aes(x, y)) + geom_point() + 
  stat_debug_group(summary.fun = function(x) {x})

With grouping, for each group the compute_group() function is called with a subset of the data.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_debug_group(summary.fun = head, summary.fun.args = list(n = 3))

## [1] "Input 'data' to 'compute_group()':"
##   x          y colour PANEL group
## 1 1 -27205.450      A     1     1
## 3 3  45790.918      A     1     1
## 5 5  -8028.578      A     1     1
## [1] "Input 'data' to 'compute_group()':"
##   x         y colour PANEL group
## 2 2 -14242.65      B     1     2
## 4 4  53731.42      B     1     2
## 6 6 102863.94      B     1     2

In this example with grouping and facets, within each panel the compute_group() function is called for each group, in total four times.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_debug_group(summary.fun = nrow) +
  facet_wrap(~block)

## [1] "Input 'data' to 'compute_group()':"
## [1] 25
## [1] "Input 'data' to 'compute_group()':"
## [1] 25
## [1] "Input 'data' to 'compute_group()':"
## [1] 25
## [1] "Input 'data' to 'compute_group()':"
## [1] 25

With facets, for each panel the compute_panel() function is called with a subset of the data that is not split by groups. For our example, it is called twice.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_debug_panel(summary.fun = nrow) +
  facet_wrap(~block)

## [1] "Input 'data' to 'compute_panel()':"
## [1] 50
## [1] "Input 'data' to 'compute_panel()':"
## [1] 50

Finally we show how geom_debug() can be used. First to print to the console the data as passed to geoms.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  geom_debug(summary.fun = head)

## Input 'data' to 'geom_debug()':

##    colour x          y PANEL group
## 1 #F8766D 1 -27205.450     1     1
## 2 #00BFC4 2 -14242.651     1     2
## 3 #F8766D 3  45790.918     1     1
## 4 #00BFC4 4  53731.420     1     2
## 5 #F8766D 5  -8028.578     1     1
## 6 #00BFC4 6 102863.943     1     2

And also to print to the console the data returned by a stat.

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_smooth(method = "lm",
             geom = "debug", 
             summary.fun = as_data_frame, 
             summary.fun.args = list())

## Input 'data' to 'geom_debug()':

## # A tibble: 160 x 8
##     colour         x          y      ymin       ymax       se PANEL group
##      <chr>     <dbl>      <dbl>     <dbl>      <dbl>    <dbl> <int> <int>
## 1  #F8766D  1.000000 -188136.10 -254710.6 -121561.62 33111.18     1     1
## 2  #F8766D  2.240506 -176933.14 -242260.5 -111605.80 32490.90     1     1
## 3  #F8766D  3.481013 -165730.18 -229819.0 -101641.36 31874.92     1     1
## 4  #F8766D  4.721519 -154527.23 -217386.7  -91667.76 31263.50     1     1
## 5  #F8766D  5.962025 -143324.27 -204964.1  -81684.46 30656.89     1     1
## 6  #F8766D  7.202532 -132121.31 -192551.7  -71690.88 30055.40     1     1
## 7  #F8766D  8.443038 -120918.36 -180150.3  -61686.39 29459.34     1     1
## 8  #F8766D  9.683544 -109715.40 -167760.5  -51670.30 28869.04     1     1
## 9  #F8766D 10.924051  -98512.44 -155383.0  -41641.90 28284.87     1     1
## 10 #F8766D 12.164557  -87309.48 -143018.6  -31600.41 27707.21     1     1
## # ... with 150 more rows

ggplot(my.data, aes(x, y, colour = group)) + geom_point() + 
  stat_peaks(span = NULL,
             geom = "debug", 
             summary.fun = as_data_frame, 
             summary.fun.args = list())

## Input 'data' to 'geom_debug()':

## # A tibble: 2 x 10
##    colour xintercept yintercept label     x       y PANEL group x.label
##     <chr>      <dbl>      <dbl> <chr> <dbl>   <dbl> <int> <int>   <chr>
## 1 #F8766D         95     984858    95    95  984858     1     1      95
## 2 #00BFC4        100    1077468   100   100 1077468     1     2     100
## # ... with 1 more variables: y.label <chr>

formula <- y ~ poly(x, 3, raw = TRUE)
ggplot(my.data, aes(x, y)) +
  stat_fit_residuals(formula = formula, 
                     geom = "debug",
                     summary.fun = as_data_frame, 
                     summary.fun.args = list())

## Input 'data' to 'geom_debug()':

## # A tibble: 100 x 6
##        x          y    y.resid y.resid.abs PANEL group
##    <dbl>      <dbl>      <dbl>       <dbl> <int> <int>
## 1      1 -23513.687 -23513.687   23513.687     1    -1
## 2      2 -11663.732 -11663.732   11663.732     1    -1
## 3      3  47289.460  47289.460   47289.460     1    -1
## 4      4  54175.231  54175.231   54175.231     1    -1
## 5      5  -8620.676  -8620.676    8620.676     1    -1
## 6      6 101247.937 101247.937  101247.937     1    -1
## 7      7 -21182.019 -21182.019   21182.019     1    -1
## 8      8   9425.407   9425.407    9425.407     1    -1
## 9      9  75240.365  75240.365   75240.365     1    -1
## 10    10 -50439.597 -50439.597   50439.597     1    -1
## # ... with 90 more rows

formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y)) +
  geom_point() +
  stat_fit_augment(method = "lm", 
                   method.args = list(formula = formula),
                   geom = "debug",
                   summary.fun = tibble::as_data_frame, 
                   summary.fun.args = list()) +
  stat_fit_augment(method = "lm", 
                   method.args = list(formula = formula),
                   geom = "smooth",
                   aes(y = ...fitted..,
                       ymax = ...fitted.. + ...se.fit.. * 2,
                       ymin = ...fitted.. - ...se.fit.. * 2))

## Input 'data' to 'geom_debug()':

## # A tibble: 100 x 17
##         ymin     ymax          y     x I.x.2. I.x.3.    .fitted  .se.fit
##        <dbl>    <dbl>      <dbl> <dbl>  <dbl>  <dbl>      <dbl>    <dbl>
## 1  -48698.27 41314.74 -3691.7638     1      1      1 -3691.7638 22673.48
## 2  -44308.35 39150.51 -2578.9186     2      4      8 -2578.9186 21022.55
## 3  -40181.60 37184.51 -1498.5419     3      9     27 -1498.5419 19487.84
## 4  -36313.71 35426.09  -443.8109     4     16     64  -443.8109 18070.62
## 5  -32700.69 33884.88   592.0973     5     25    125   592.0973 16772.32
## 6  -29338.59 32570.60  1616.0056     6     36    216  1616.0056 15594.38
## 7  -26223.08 31492.56  2634.7369     7     49    343  2634.7369 14538.06
## 8  -23348.80 30659.03  3655.1141     8     64    512  3655.1141 13604.10
## 9  -20708.60 30076.52  4683.9600     9     81    729  4683.9600 12792.32
## 10 -18292.59 29748.78  5728.0975    10    100   1000  5728.0975 12101.20
## # ... with 90 more rows, and 9 more variables: .resid <dbl>, .hat <dbl>,
## #   .sigma <dbl>, .cooksd <dbl>, .std.resid <dbl>, y.observed <dbl>,
## #   t.value <dbl>, PANEL <int>, group <int>

formula <- y ~ x + I(x^2) + I(x^3)
ggplot(my.data, aes(x, y2, colour = group)) +
  geom_point() +
  stat_fit_augment(method = "lm", 
                   method.args = list(formula = formula),
                   geom = "debug",
                   summary.fun = tibble::as_data_frame, 
                   summary.fun.args = list()) +
  stat_fit_augment(method = "lm", 
                   method.args = list(formula = formula),
                   geom = "smooth",
                   aes(y = ...fitted..,
                       ymax = ...fitted.. + ...se.fit.. * 2,
                       ymin = ...fitted.. - ...se.fit.. * 2))

## Warning: Computation failed in `stat_fit_augment()`:
## variable lengths differ (found for 'x')

## Input 'data' to 'geom_debug()':

## # A tibble: 100 x 19
##     colour      ymin     ymax .rownames           y     x I.x.2. I.x.3.
##      <chr>     <dbl>    <dbl>     <chr>       <dbl> <dbl>  <dbl>  <dbl>
## 1  #F8766D -41080.41 20399.43         1 -10340.4868     1      1      1
## 2  #F8766D -35273.92 17554.72         3  -8859.5995     3      9     27
## 3  #F8766D -30156.45 15352.63         5  -7401.9119     5     25    125
## 4  #F8766D -25719.57 13832.64         7  -5943.4676     7     49    343
## 5  #F8766D -21946.27 13025.65         9  -4460.3100     9     81    729
## 6  #F8766D -18793.72 12936.75        11  -2928.4828    11    121   1331
## 7  #F8766D -16175.63 13527.57        13  -1324.0295    13    169   2197
## 8  #F8766D -13958.28 14712.29        15    377.0064    15    225   3375
## 9  #F8766D -11978.53 16375.69        17   2198.5812    17    289   4913
## 10 #F8766D -10073.31 18402.61        19   4164.6515    19    361   6859
## # ... with 90 more rows, and 11 more variables: .fitted <dbl>,
## #   .se.fit <dbl>, .resid <dbl>, .hat <dbl>, .sigma <dbl>, .cooksd <dbl>,
## #   .std.resid <dbl>, y.observed <dbl>, t.value <dbl>, PANEL <int>,
## #   group <int>

The package also defines a "null" geom, which is used as default by the debug stats described above. Currently this geom is similar to the recently added ggplot2::geom_blank(), which should be preferred.

ggplot(my.data, aes(x, y, colour = group)) + geom_null()