1 #!/usr/bin/python2.7 2 # Copyright (c) 2014 The Chromium OS Authors. All rights reserved. 3 # Use of this source code is governed by a BSD-style license that can be 4 # found in the LICENSE file. 5 6 import numpy 7 8 9 def LinearRegression(x, y): 10 """Perform a linear regression using numpy. 11 12 @param x: Array of x-coordinates of the samples 13 @param y: Array of y-coordinates of the samples 14 @return: ((slope, intercept), r-squared) 15 """ 16 # p(x) = p[0]*x**1 + p[1] 17 p, (residual,) = numpy.polyfit(x, y, 1, full=True)[:2] 18 # Calculate the coefficient of determination (R-squared) from the 19 # "residual sum of squares" 20 # Reference: 21 # http://en.wikipedia.org/wiki/Coefficient_of_determination 22 r2 = 1 - (residual / (y.size*y.var())) 23 24 # Alternate calculation for R-squared: 25 # 26 # Calculate the coefficient of determination (R-squared) as the 27 # square of the sample correlation coefficient, 28 # which can be calculated from the variances and covariances. 29 # Reference: 30 # http://en.wikipedia.org/wiki/Correlation#Pearson.27s_product-moment_coefficient 31 #V = numpy.cov(x, y, ddof=0) 32 #r2 = (V[0,1]*V[1,0]) / (V[0,0]*V[1,1]) 33 34 return p, r2 35 36 37 def FactsToNumpyArray(facts, dtype): 38 """Convert "facts" (list of dicts) to a numpy array. 39 40 @param facts: A list of dicts. Each dict must have keys matching the field 41 names in dtype. 42 @param dtype: A numpy.dtype used to fill the array from facts. The dtype 43 must be a "structured array". ie: 44 numpy.dtype([('loops', numpy.int), ('cycles', numpy.int)]) 45 @returns: A numpy.ndarray with dtype=dtype filled with facts. 46 """ 47 a = numpy.zeros(len(facts), dtype=dtype) 48 for i, f in enumerate(facts): 49 a[i] = tuple(f[n] for n in dtype.names) 50 return a 51
