Skip to content

errata.md

Latest commit

 

History

History
32 lines (29 loc) · 4.81 KB

errata.md

File metadata and controls

32 lines (29 loc) · 4.81 KB
Page Printed text Correct text Thanks
1 We can summarize the Bayesian modeling process using three steps Missing citation of Bayesian Data Analysis by Gelman et al Bob Carpenter
5 So let’s take a walk through the garden of forking paths [Borges, 1944]. In the context of statistics the "garden of forking paths" appears mentioned By Andrew Gelman, Richard McElreath in his book Statistical Rethinking, see also this wikipedia entry Bob Carpenter
13 The binomial coefficient is typeset as $\left(\frac{n}{x}\right)$ it should be $\left(n\atop{}x\right)$ Chris Hansen
32 There is a typo in the denominator of the normalizing constant it should be $p(\theta) = \underbrace{\frac{\Gamma(\alpha + \beta)}{\Gamma(\alpha) \Gamma(\beta)}}_{\text{normalizing constant}} \theta^{\alpha-1} (1-\theta)^{\beta-1}$ XIN Hongwei
62 We will explore the value of over a grid of 200 points. We will explore the value over a grid of 200 points. dweights
64 ...and many operations applied to Guassians return another Gaussian. ...and many operations applied to Gaussians return another Gaussian. marctagl65
88 (exercise 4) wrong short url The link should point to https://www.pymc.io/projects/docs/en/stable/learn/core_notebooks/pymc_overview.html#case-study-2-coal-mining-disasters DrEntropy
94 ...the variance between observed and theoretical values should be the same for all groups. ...the variance between observed and theoretical values should be unique for each group. Kenji Oman
104 Figure 3.6 indicates a HalfNormal Figure 3.6 should indicate a Gamma Jacob Warren
115 We are going to usetemperature We are going to use temperature Kenji Oman
115 The noise term is 𝜖 The noise term is 𝜎 Parrenin Frédéric
116 We commit it because otherwise... We omit it because otherwise... Kenji Oman
122 The variance of the NegativeBinomial is 𝜇 + 𝜇²/𝛼 , so the larger the value of 𝛼 the larger the variance. The variance of the NegativeBinomial is 𝜇 + 𝜇²/𝛼 , so the larger the value of 𝛼 the smaller the variance. Tomás Capretto
124 (or data with a few bulk points) (or data with only a few bulk points) Kenji Oman
133 We have been using the linear motif to model the mean of a distribution and, in the previous section, we used it to model interactions. In statistics,... We have been using the linear motif to model the mean of a distribution. In statistics,... Jacob Warren
145 In the next chapter, we will learn more about linear regression... In Chapter 6, we will learn more about linear regression... Tomás Capretto
155 In the equation for a polynomial model of order 5 all coefficient are the same ($\beta_{0}$) it should be $\alpha + \beta_0 x + \beta_1 x^2 + \beta_2 x^3 + \beta_3 x^4 + \beta_4 x^5$ Jarvin Jeffrey Gallego
191 The utility of plot_cap ... The utility of plot_predictions... Tomás Capretto
194 model_poly4 = bmb.Model("rented ∼ poly(temperature, degree=4)", bikes, model_poly4 = bmb.Model("rented ∼ poly(hour, degree=4)", bikes, Jacob Warren
195 Figure 6.5: Posterior mean and posterior predictive distribution for the bikes model with temperature and humidity Figure 6.5: Posterior mean and posterior predictive distribution for the polynomial bikes models with hour. Jacob Warren
207 We have been using bmb.interpret_plot_predictions ... One of them is bmb.interpret_plot_comparisons. We have been using bmb.interpret.plot_predictions ... One of them is bmb.interpret.plot_comparisons. Tomás Capretto
208 Another useful function is bmb.interpret_plot_slopes Another useful function is bmb.interpret.plot_slopes Tomás Capretto
254 We call 𝜙 the inverse link function and 𝜙 is... We call 𝜓 the inverse link function and 𝜙 is... Jacob Warren
344 https://arviz-devs.github.io/Exploratory-Analysis-of-Bayesian-Models/ https://arviz-devs.github.io/EABM

Notes:

  • On page 23, Figure 1.9 shows the kurtosis. What PreliZ actually computes is the "excess kurtosis", i.e the kurtosis -3. Thanks to Narinder Singh for pointing this out.

  • On Code block 2.26 we use a Normal likelihood and in Code block 2.27 we use a Gamma likelihood. The text has been updated to reflect why we do the change (The Gamma likelihood is more appropriate for the data we are modeling, given that all values are positive). Thanks to Kenji Oman for pointing this out.