Numpy

• import numpy as np
• array: x = np.array([3, 4, 5])
  • adding arrays together works as expected
  • arranged as a sequence of rows
• matrix
  • np.matrix()
  • x = np.array([[1, 2], [3, 4]])
  • Uses [0,0] indexing: x[row,col]
  • x[2] yields the third entry of x
  • correlation matrix: np.corrcoef(x, y)
• attributes: x.attribute
  • dimensions: x.ndim
  • data type: x.dtype
  • nrows and ncols: x.shape
  • transpose: x.T
• methods: x.method()
  • function associated with an object
  • sum: x.sum() or np.sum(x)
  • reshape: x.reshape(tuple)
    • returns a new array with the elements of x in a different shape
    • modifying values in a reshaped object also changes the values in the original object
  • square root: np.sqrt(x)
  • squaring: x**2
  • mean: np.mean()
    • axis=0 refers to matrix rows
  • variance: np.var()
    • default: divides by $n$ instead of $n-1$
    • axis=0 refers to matrix rows
  • standard deviation: np.std()
    • axis=0 refers to matrix rows
• generate random data: np.random.normal()
  • set a seed: x=np.random.default_rng() then x.normal()
• pi: np.pi
• np.isnan() finds NA values

Sequences and Slice Notation

• vector of $n$ numbers starting at $a$ and ending at $b$: np.linspace(a,b,n)
• vector of numbers spaced out by step starting at $a$ and ending at $b$: np.arange(a,b)
• slice: [a:b] i.e. "hello world"[3:6] outputs 'lo '

Indexing Data

• Select multiple rows: A[[1,3]]
• Select multiple columns: A[:,[0,2]]
• Select submatrix
  • function: A[np.ix_([1,3],[0,2,3])]
  • slices: A[1:4:2,0:3:2]
• Booleans
  • np.zeros() make an array with all zeros (set dtype=bool)
  • use "= =" w/o space to verify if two objects are equal
  • check all entries are true: np.all()
  • check if any entries are true: np.any()
  • boolean arrays can be used to select submatrices while equivalent binary arrays can't
• np.argmax() identifies the index of the largest element of an array, optionally computed over an axis of the array
• df.corr() to see correlation matrix
• use np.set_printoptions(threshold=sys.maxsize) to stop truncating outputs and np.set_printoptions(threshold=False) to return to truncating results
• df.describe() to get summary data for each numeric column