statswithr / book Goto Github PK

While explaining a Bayesian approach it says:
"Before she saw the data, the Bayesian’s uncertainty expressed by her standard deviation was 0.71. After seeing the data, it was much reduced – her posterior standard deviation is just 0.13."

But, considering beta(1,1) before the data and beta(1,5) after the data, the standard deviation should be 0.289 and 0.141 respectively.

At the end of chapter 3.1, the last sentence is:

And in decision theory, one seeks to minimize one's expected loss

Coming from an economics perspective, we seek to maximise gain. from a maths perspective minimising F(x) is the same as maximising G(x) = -F(x). Is this worth a footnote?

Then in section 3.2, paragraph four states:

Here, of course, instead of minimizing expected losses, we want to maximize the expected gain.

Then it is suggested to use the mode. I think the point is that whether the problem is max gain or min loss makes no difference - in this case the binary loss (or gain) function is why the mode is now relevant. Perhaps this could be made clearer?

Ps - I found this diagram helpful to clarify for me.
https://www.stat.auckland.ac.nz/~brewer/stats331.pdf

Please check the library 'statsr'. I failed to compile the book due to this issue.

In section 1.1.2, the equations are cut off:

Not sure if this extends to other areas.

At the end of section 1.1.2, the same question is asked to the reader twice.

If an individual is at a higher risk for having HIV than a randomly sampled person from the population considered, how, if at all, would you expect P(Person tested has HIV∣ELISA is positive) to change?
...
Example 1.3 If the an individual is at a higher risk for having HIV than a randomly sampled person from the population considered, how, if at all, would you expect P(Person tested has HIV ∣ ELISA is positive) to change?
...

If you agree with this issue, I'm happy to fork and send a PR to fix it.

Great book! I read the pdf version. As html is probably your primary version, it could be a low priority to have a “beautiful” pdf version, so feel free to ignore these suggestions.

1. If page numbers were consistently at the bottom, it would give you more space for the sometimes long titles, such as page 108. To fix, see E.g. https://community.rstudio.com/t/custom-position-of-page-numbers/7873?u=mbn

2. If you are happy that chapter is redundant, that would also give you more room for text on page headers. To fix, see E.g. https://stackoverflow.com/questions/43757899/how-to-remove-the-word-chapter-from-the-chapter-headings

3. Figures (and tables) are sometimes printed well before they are mentioned in the text. I think this happens automatically by default. My personal preference would be that figures are always after they are mentioned, even if it ends up creating more white space. To fix, see eg https://community.rstudio.com/t/cant-control-position-of-tables-and-figures-in-knitted-pdf-document/37364/3?u=mbn

4. Code occasionally goes a bit wide e.g. p.126. However, most people that want to use the code will be looking at an electronic version, so maybe this doesn’t matter?

5. Some people may print in black and white, in which case some graphs will still be legible (e.g. Figure 5.1 with dots and dashes), and others will not (e.g. Figure 2.4).

In Section 8.4 the coefficients that are extracted from the coefficients object do not have zeros where expected for the Best predictive model due to an issue that the reference to the best model refers to the location of the best model in the original bas object but the order to the output from coef.bas is sorted based on model probabilities. Updated the chapter with the correct code so that the coefficients correspond to the correct model. See Issue merliseclyde/BAS#49

On the book's website, section "4.1.7.1 Derivation of Marginal Distribution for μ" (see https://statswithr.github.io/book/inference-and-decision-making-with-multiple-parameters.html#sec:normal-gamma) shows "\intertext" in red, and the included text is cropped.

Note: The amsmath package was once in preamble.tex, but commit 600c9db removed it.