Feature #18057: Introduce Array#mean

• Status: Open
• Assignee: -
• Target version: -

[ggmichaelgo (Michael Go)]

Introduce Array#average to calculate the average value of an array.

array = [1, 2, 3]
array.mean # 2

array = [1.5, 2.2, 3.1]
array.mean(&:round) # 2.3333333333333335

array = [-3, -2, -1]
array.mean { |e| e.abs } # 2

[ggmichaelgo (Michael Go)]

I have created PR for this feature :)

https://github.com/ruby/ruby/pull/4699

[sawa (Tsuyoshi Sawada)]

The proposal is unclear. It seems to be neither arithmetic nor geometric mean. What kind of average do you have in mind? What is the reason you think that kind of average is particularly needed? What does the block do?

[ggmichaelgo (Michael Go)]

Hello Sawa! I apologize for the confusion... (this is my first time contributing, and I am hoping to contribute more :D)

For this function, I am trying to calculate the arithmetic mean.
I have been developing a stock price analytics algorithm, and I would be using this function to calculate the moving average price.

ref on Moving Average

The block is used like the rb_ary_sum, and it is use to calculate the sum of the array.

[sawa (Tsuyoshi Sawada)]

ggmichaelgo, thanks for the reply. I have thought that what you wrote as the comment in the example code snippet are the expected return values. You made clear in note#3 that they are not. Then, what are they representing? If they are random comments irrelevant to the proposal, then you should delete them to avoid confusion.

[ggmichaelgo (Michael Go)]

• Description updated (diff)

[ggmichaelgo (Michael Go)]

oh goodness... that is embarrassing... Thank you for catching my error... 🙏

[ggmichaelgo (Michael Go)]

• Description updated (diff)

[mrkn (Kenta Murata)]

I created enumerable-statistics gem to provide the methods for calculating precise statistical summary in Array and Enumerable. Does this gem fit your use cases?

I think this is unnecessary to provide in the core because average values can be calculated by ary.sum / ary.length without floating-point arithmetic errors. If we provide this method in the core, why not provide stddev or variance, which need complex calculations to avoid floating-point arithmetic errors.

If this feature request will be accepted by any chance, I prefer mean over average because in many languages, such a functions are named mean instead of average. For example Python's statistics.mean, pandas's pandas.Series.mean, NumPy's numpy.mean, Julia's Statistics.mean, Boost's <boost/accumulators/statistics/mean.hpp>, etc.

[ggmichaelgo (Michael Go)]

• Subject changed from Introduce Array#average to Introduce Array#mean
• Description updated (diff)

I definitely prefer using mean over average!

[ggmichaelgo (Michael Go)]

Thank you @mrkn (Kenta Murata)!
I didn't know about enumerable-statistics, and I can definitely use this in my projects!!
I wish I knew about it sooner!

[bkuhlmann (Brooke Kuhlmann)]

💡 You can also use the Refinements gem which also provides #mean support.

[nobu (Nobuyoshi Nakada)]

When ary has only Integers, ary.sum is also an Integer and ary.sum / ary.length too.

[1,2].sum     #=> 3
[1,2].sum / 2 #=> 1

I don't think this is the expected result, no?
So it should be ary.sum.quo(ary.length) or ary.sum.fdiv(ary.length).

[ggmichaelgo (Michael Go)]

@nobu (Nobuyoshi Nakada) - yup, you are correct. The expected behaviour should be like this

[1,2].mean #=> 1.5

[matheusrich (Matheus Richard)]

Was this proposal rejected?