What is the proper way plot multiple lambda function in plotnine so legend appears?

Question

This looks like it's fairly close to the answer I'm looking for but it's not quite there. Perhaps I'm not following the "tidy data" principles and will need to make a dataframe evaluated at many points to plot these functions but I'm hesitant to accept that as the answer.

Here's the code to plot the graph I have.

call_value_per_unit = lambda s_t1,X: max(0, s_t1-X)
put_value_per_unit = lambda s_t1, X: max(0, X-s_t1)
put_call_value = lambda s_t1, X: put_value_per_unit(s_t1, X) + call_value_per_unit(s_t1, X)

independent_variable = "Stock Price"
dependent_variable = "Asset Price"

g = ggplot(pd.DataFrame({independent_variable:[10,20]}), aes(x=independent_variable)) \
         + stat_function(fun=put_value_per_unit, args=[15], color="red") \
         + stat_function(fun=call_value_per_unit, args=[15], color="blue") \
         + stat_function(fun=put_call_value, args=[15], color="black") \
         + ylab(dependent_variable) \
         + ggtitle(" ".join([independent_variable , "vs", dependent_variable]))
_ = g.draw()

But there's no legend... And I'd like there to be one.

(Although I'm in python, R users will likely have good suggestions)

which field you want as legend and what kind of plot are you expecting? — nikn8, Apr 20 '20 at 08:22
It's fairly easy to plot with tidy approach. are you ok with that? — nikn8, Apr 20 '20 at 08:27
I suppose. To make sure I understand the tidy approach, I'd make a dataframe with these functions evaluated at many points, and then plot them using geom_line? — financial_physician, Apr 20 '20 at 08:30
R programmer here. Shouldn't there be a `geom_*` like `geom_point`? Or, in `stat_function`, a `stat = "point"` or `"line"` or other value? — Rui Barradas, Apr 20 '20 at 08:41
@RuiBarradas You are right. There is nothing like stat_function, and as you mentioned it's either geom_* or stat_* (e.g.: stat_count(geom = "bar") .. — nikn8, Apr 20 '20 at 08:48
@financial_physician I understand, but output of the functions should be part of data frame or it could be complex to plot. would you be able to share sample data? — nikn8, Apr 20 '20 at 08:51
@Neel I'm actually not working with any data. I'm just trying to make some pretty graphs and ggplot has been good for data plots. Maybe I should use a different tool since I'm not working with a dataset right now — financial_physician, Apr 20 '20 at 09:15
@RuiBarradas what do R programmers usually use if they just want to plot some functions on a graph? — financial_physician, Apr 20 '20 at 09:17

score 1 · Answer 1 · answered Apr 20 '20 at 11:29

1

In R, to plot functions with package ggplot2, first define a data set with a x vector. Then use stat_function with an appropriate geom. This usually is one of

geom = "line"
geom = "point"

Then, it's very simple to graph the function.

library(ggplot2)

df1 <- data.frame(x = -5:5)

ggplot(df1, aes(x)) +
  stat_function(geom = "line", fun = function(x) x^2)

answered Apr 20 '20 at 11:29

Rui Barradas

70,273
8
34
66

Thanks Rul! Really appreciate the insight – financial_physician Apr 20 '20 at 21:35

score 1 · Accepted Answer · answered Apr 20 '20 at 14:32

Plotting multiple functions in one plot and having a legend for each function is not possible. For stat_function, func is a parameter not an aesthetic so you cannot map a variable/column to it. Legends only help interpret aesthetic mappings.

Since you want to do heavy many computations, do that outside the plotting calls then plot the results with geom_line. Make sure your dataframe is in tidy data form. Do not let the fact that there is a stat_function force you into using when it is not the best tool for the job.

I think this is the best answer, it walks through what's wrong with what I want to do, why I should do it differently, and then how to go about it differently. Thank you! — financial_physician, Apr 20 '20 at 21:36

What is the proper way plot multiple lambda function in plotnine so legend appears?

2 Answers2