Errata

The errata list is a list of errors and their corrections that were found after the product was released. Use the Github issue tracker to submit new errors.

Chapter 1

• p. 12-13, code for Figure 1.7 and Figure 1.8 - value_counts creates now column named proportion:

  ...
  X.loc[4, 'proportion'] = 0  # there are no samples with 4 blemishes add a row
  X = X.sort_index()  # sort by number of blemishes
  ax = X['proportion'].plot.bar(color='grey', legend=False)
  ...
  X['Cumulative Frequency'] = X['proportion'].cumsum()
  ...
  

Chapter 2

• p. 55, Section 2.2.1.1 - Clarification: The scipy package uses the term probability mass function instead of probability distribution function for discrete distributions. Both terms are used in the literature. In Modern Statistics, we use the term probability distribution function for both discrete and continuous distributions.

Chapter 3

• p. 162, Equation 3.30 - Replace percentile for the first $\chi^2$:

  $\left(\frac{(n-1)S^2}{\chi^2{1-\alpha/2}[n-1]}, \frac{(n-1)S^2}{\chi^2{\alpha/2}[n-1]}\right)$

• p. 186, Figure 3.15 - Incorrect distribution used to create the Figure. The updated code creates the following Figure. The additional dashed lines show the approximated HPD.
• p. 190 - Code sample, the compute_bootci function cannot identify np.mean as a function (see PR). Replace with

  B = pg.compute_bootci(etchrate, func=lambda x: np.mean(x), n_boot=1000,
                      confidence=0.95, return_dist=True, seed=1)
  

• p. 196 - Code sample, the print statement needs to contain f-strings. Replace

  print('Xbar {Xbar:.2f} / SX {SX:.3f}')
  print('Ybar {Xbar:.2f} / SY {SX:.3f}')
  

  with

  print(f'Xbar {Xbar:.2f} / SX {SX:.3f}')
  print(f'Ybar {Ybar:.2f} / SY {SY:.3f}')
  

Chapter 4

• p. 251: Scale in the regression summary output corresponds to $S_e^2$.

  The $S_e^2$ value is shown in the regression summary output as Scale.

• Exercise 4.27: effects => affects

Chapter 6

• p. 333 - naming inconsistency of formula for polynomial model with seasonality. Replace poly_formula with formula:

  formula = 'Dow_Jones_Index ~ C(month) + I(trend**3) + I(trend**2) + trend + 1'
  model_3 = smf.ols(formula=formula, data=dow1941_df).fit()
  

• p. 357 - Correct value of PMSE

  The empirical PMSE is 0.8505.

Chapter 7

• p. 379 - iteritems deprecated in pandas package; code change required:

  df = pd.DataFrame([
    {satisfaction: counts for satisfaction, counts
      in response.value_counts().items()},
    {satisfaction: counts for satisfaction, counts
      in response[q1_5].value_counts().items()},
  ])
  

Chapter 8

• p. 400 - Import from from skfda.preprocessing.dim_reduction.projection is deprecated. Use

  from skfda.preprocessing.dim_reduction import FPCA
  

• p. 404 - In equation 8.2, the matrix $V$ needs to be transposed:
  $DTM \approx U * S * V’$
• p. 408 - estimating the network using HillClimbSearch now requires setting max_indegree to 1

  est = HillClimbSearch(data=abc)
  model = est.estimate(max_indegree=1, max_iter=int(1e4), show_progress=False,
                      scoring_method='k2score')
  

  the updated code creates the following Figure 8.11 on p. 414 and requires adjustment of the text in Example 8.4

  We can see that changes to the Equipment and Suppliers have the largest ATE.

Index

• P-value, 152, 215