Order from chaos

This part will cover

Sorting / grade up / grade down
Windowed reduce

Your training as a climate sceintist has prepared you to deal with measurement of real-life systems, which can have a large amount of noise that has to be dealt with before making any inferences. Trying to test your data reduction methods on the computer, you first generate artificially noisy data.

The monadic ? roll function generates a random number for every scalar in its right argument, from 1 to the value of that scalar.

       ⍝ Roll a 6-sided die
       ? 6
2
       ? 6 6 6
6 5 3
       ? ⍳10
1 1 3 4 4 6 6 5 8 4

       ? 10*⍳10
8 84 545 4600 99210 693088 7548981 55682236 454935262 2471243355

When the argument contains zero (0), a random floating point number between 0 and 1 is returned.

       ? 0 0 0 0
0.5073955156 0.1942971103 0.3645209613 0.8563507943
       ? 0,⍳4
0.1770380776 1 1 2 4

The dyadic ? deal function works similarly, but generates a list of random numbers without repeats.

      5 ? 5
2 5 4 1 3
      5 ? 10
5 7 1 6 8
      10 ? 10
7 5 6 10 1 9 4 2 8 3
      11 ? 10
DOMAIN ERROR: Deal right argument must be greater than or equal to the left argument
      11 ? 10
         ∧

The replicate function can be used to generate a large number of random number in a certain range, by creating a vector of the same number of arbitrary length.

       6 / 10
10 10 10 10 10 10 10
       ? 6 / 10
9 9 2 9 4 6

Simulating measuring a temperature of 21 degrees with some noise, fluctuating between -0.5 and 0.5

       DATA ← 21 + (? 20 / 0) - 0.5
20.51270976 20.84443059 21.20961766 20.85869017 21.05525209 20.8088396 21.00197974 20.88101536 20.70209893 21.28923291
      20.62552888 20.95059287 20.78032757 20.52831086 21.29706401 21.45739671 20.94876038 20.84444697 20.66278673
      20.76248039

It is possible to recover the original data by taking an average, using the reduce / operator on + plus then dividing by the ≢ tally of the DATA vector.

       (+/DATA)÷(≢DATA)
20.90107811
       (+/÷≢) DATA
20.90107811

For measurements of time-varying data, it would be useful instead to do a moving average instead. Take, for example, a simulated temperature reading of air temperature that goes from ¯20 to ¯30 over the course of a day's measurements.

       DATA ← ¯20-(1÷10)×⍳100
¯20.1 ¯20.2 ¯20.3 ¯20.4 ¯20.5 ¯20.6 ¯20.7 ¯20.8 ¯20.9 ¯21 ¯21.1 ¯21.2
 ¯21.3 ¯21.4 ¯21.5 ¯21.6 ¯21.7 ¯21.8 ¯21.9 ¯22 ¯22.1 ¯22.2 ¯22.3 ¯22.4
 ¯22.5 ¯22.6 ¯22.7 ¯22.8 ¯22.9 ¯23 ¯23.1 ¯23.2 ¯23.3 ¯23.4 ¯23.5 ¯23.6
 ¯23.7 ¯23.8 ¯23.9 ¯24 ¯24.1 ¯24.2 ¯24.3 ¯24.4 ¯24.5 ¯24.6 ¯24.7 ¯24.8
 ¯24.9 ¯25 ¯25.1 ¯25.2 ¯25.3 ¯25.4 ¯25.5 ¯25.6 ¯25.7 ¯25.8 ¯25.9 ¯26
 ¯26.1 ¯26.2 ¯26.3 ¯26.4 ¯26.5 ¯26.6 ¯26.7 ¯26.8 ¯26.9 ¯27 ¯27.1 ¯27.2
 ¯27.3 ¯27.4 ¯27.5 ¯27.6 ¯27.7 ¯27.8 ¯27.9 ¯28 ¯28.1 ¯28.2 ¯28.3 ¯28.4
 ¯28.5 ¯28.6 ¯28.7 ¯28.8 ¯28.9 ¯29 ¯29.1 ¯29.2 ¯29.3 ¯29.4 ¯29.5 ¯29.6
 ¯29.7 ¯29.8 ¯29.9 ¯30


      ⍝ Add noise
      DATA +← (? 100 / 0) - 0.5

In order to do a moving average, a function that takes moving windows of a vector is needed, the windowed reduce / operator is exactly that. It applies the reduce function on moving segments of its right argument specific by its left argument.

       3 ,/ ⍳10
┌─────┬─────┬─────┬─────┬─────┬─────┬─────┬──────┐
│1 2 3│2 3 4│3 4 5│4 5 6│5 6 7│6 7 8│7 8 9│8 9 10│
└─────┴─────┴─────┴─────┴─────┴─────┴─────┴──────┘

       3 +/ ⍳10
6 9 12 15 18 21 24 27
       (1+2+3) (2+3+4) (3+4+5) (4+5+6) (5+6+7) (6+7+8) (7+8+9) (8+9+10)
6 9 12 15 18 21 24 27

       2 ×/ ⍳10
2 6 12 20 30 42 56 72 90
       (1×2) (2×3) (3×4) (4×5) (5×6) (6×7) (7×8) (8×9) (9×10)
2 6 12 20 30 42 56 72 90

       3 3 ⍴ ⍳9
1 2 3
4 5 6
7 8 9
       2,/(3 3 ⍴ ⍳9)
┌───┬───┐
│1 2│2 3│
├───┼───┤
│4 5│5 6│
├───┼───┤
│7 8│8 9│
└───┴───┘
      2+/(3 3 ⍴ ⍳9)
 3  5
 9 11
15 17

To get a moving average, the windowed reduce function can be applied to the + plus function and then divided by the size of the window, in this case 5. The result is then rounded to 1 decimal place.

      RESULT ← (5+/DATA)÷5
      (⌈RESULT×10)÷10
¯20.4 ¯20.5 ¯20.7 ¯20.7 ¯20.8 ¯20.8 ¯20.9 ¯20.9 ¯21.1 ¯21.1 ¯21.2 ¯21.4
 ¯21.5 ¯21.6 ¯21.8 ¯21.9 ¯21.9 ¯22.1 ¯22.2 ¯22.3 ¯22.5 ¯22.7 ¯22.7 ¯22.8
 ¯22.8 ¯22.8 ¯22.8 ¯22.9 ¯23 ¯23.1 ¯23.2 ¯23.5 ¯23.7 ¯23.6 ¯23.8 ¯23.9 ¯24
 ¯24.1 ¯24.4 ¯24.5 ¯24.6 ¯24.5 ¯24.5 ¯24.5 ¯24.5 ¯24.5 ¯24.7 ¯24.9 ¯24.9
 ¯25.2 ¯25.4 ¯25.6 ¯25.7 ¯25.8 ¯25.8 ¯25.9 ¯25.9 ¯26.1 ¯26.2 ¯26.3 ¯26.4
 ¯26.6 ¯26.7 ¯26.7 ¯26.9 ¯27 ¯27.1 ¯27.2 ¯27.3 ¯27.3 ¯27.4 ¯27.4 ¯27.4 
 ¯27.6 ¯27.8 ¯27.9 ¯28 ¯28.2 ¯28.3 ¯28.3 ¯28.3 ¯28.4 ¯28.6 ¯28.6 ¯28.8
 ¯29 ¯29.1 ¯29.2 ¯29.3 ¯29.4 ¯29.3 ¯29.3 ¯29.4 ¯29.6 ¯29.7 ¯29.8

A pretty good reconstruction of the data!

Given the large number of temperature readings, it is also possible to sort ascending and descending using the grade up ⍋ and grade down ⍒ functions.

The grade up ⍋ and grade down ⍒ functions return indices of elements of its right argument in ascending or descending order.

       DATA ← 20 19 21 22
       ⍋DATA
2 1 3 4
       DATA[⍋DATA]
19 20 21 22
       ⍒DATA
4 3 1 2
       DATA[⍒DATA]
22 21 20 19