Make geom_smooth() from ggplot2 adapt to number of observations - r

Question

How can I make the method argument of geom_smooth() from ggplot be dynamic and adapt to the number of data points in a group?
For example, I have data in the following format:
1. DATE PRODUCT SIZE
2. 3/1/2017 A 10
3. 3/2/2017 B 14
4. 3/3/2017 C 25
5. 3/4/2017 A 16
6. etc.
This charts completely fine and adds a loess fit to each group (PRODUCT) with the following code (each PRODUCT group has about 20 entries):
DT<-read.csv("TEST_DATA.csv")
DT$DATE<-as.Date(DT$DATE, "%m/%d/%Y")
myPlot<-ggplot(DT, aes(DATE, SIZE, color = PRODUCT))
myPlot + geom_point() + geom_smooth(method = "loess", se = FALSE)
However, let's say I add in just 2 data points for a 4th Product "D". I then get the following warning messages and no loess fit lines are added to the plot for ANY group.
Warning messages:
1: In simpleLoess(y, x, w, span, degree = degree, parametric = parametric, ... : span too small. fewer data values than degrees of freedom.
I believe this warning is due to the fact that the number of observations for product D is less that the degrees of freedom for the loess fit.
Setting method = "auto", chooses "loess" anyway so that doesn't help and setting method to "lm" is not what I want.
I would like to do the following but can't quite get it to work and am wondering if someone can help?
myPlot + geom_point() + geom_smooth(data = DT, method = if(length(DT$PRODUCT)<5) {"lm"} else {"loess"}, se = F)
As you can see, I am trying to have geom_smooth() use method = "lm" if any groups have less than 5 observations, otherwise use the "loess" method. But I can't quite figure out how to access the number of observations of each group within the geom_smooth() function.

Answer 1

There's an n argument (number of points to evaluate smoother at) that you can use. See stat_smooth for details.
EDIT:
You can build the plot dynamically:
sProduct <- unique(DT$PRODUCT)
myPlot <- ggplot(DT, aes(DATE, SIZE, color = PRODUCT)) + geom_point()
for (i in sProduct){
sMethod <- ifelse(sum(DT$PRODUCT == i) <= 5, "lm", "loess")
myPlot <- myPlot + geom_smooth(data = subset(DT, PRODUCT == i), method = sMethod, se = FALSE)
}
myPlot

Answer 2

You could write a function that chooses the smoothing method conditionally, based on minimum group length. For example:
library(tidyverse)
theme_set(theme_classic())
conditional_smooth = function(data, xvar, yvar, group) {
p = ggplot(data, aes_string(xvar, yvar, colour=group)) +
geom_point()
min_group_length = split(data, data[, group]) %>% map_dbl(nrow) %>% min
# Choose smoothing method based on minimum group length
if(min_group_length >= 5) {
p + geom_smooth(method=loess)
}
else {
p + geom_smooth(method=lm)
}
}
Let's run the function. For the iris data frame, the smallest group has length 50.
conditional_smooth(iris, "Petal.Length", "Sepal.Length", "Species")
Now let's shorten one group to four values:
conditional_smooth(iris[c(1:50,97:150), ], "Petal.Length", "Sepal.Length", "Species")