In their November 3, 2018 issue, The Economist published the following figure showing their analysis of data from YouGov.

For the second edition of Think Bayes, I plan to use it to demonstrate conditional probability. In the following notebook, I replicate their analysis (loosely!) using data from the General Social Survey (GSS).

All of my books were written in LaTeX. For a long time I used emacs to compose, pdflatex to convert to PDF, Hevea to convert to HTML, and a hacked-up version of plasTeX to convert to DocBook, which is one of the formats I can submit to my publisher, O’Reilly Media.

Recently I switched from emacs to Texmaker for composition, and I recommend it strongly. I also use Overleaf for shared LaTeX documents, and I can recommend that, too.

However, the rest of the tools I use are pretty clunky. The HTML I get from Hevea is not great, and my hacked version of plasTeX is just awful (which is not plasTeX’s fault).

Since I am starting some new book projects, I decided to rethink my tools. So I asked Twitter, “If you were starting a new book project today, what typesetting language / development environment would you use? LaTeX with Texmaker? Bookdown with RStudio? Jupyter?Other?”

Nelis Willers “wrote a 510 page book with LaTeX, using WinEdt and MiKTeX and CorelDraw for diagrams. Worked really well.”

Matt Boelkins likes “PreTeXt, hands down: It has LaTeX and HTML as potential outputs among many. See the gallery of existing texts on the linked page.”

makusu recommends “Emacs org-mode. Easy to just write your content, seamless integration with latex, easy output to latex, PDF, markdown and HTML.”

AsciiDoc

Luciano Ramalho recommends “AsciiDoc, for sure. That’s how I wrote @fluentpython. It’s syntax more user-friendly than ReStructuredText and way more expressive than Markdown. AsciiDoc was *designed for* book publishing. It’s as expressive as DocBook, but it ain’t XML. With @asciidoctor you can render locally.”

JD Long provides a useful reminder: “It’s dependent on the publisher as well as the content of the book. I like Bookdown for R, but if I were doing a devops book for O’Reilly I’d write directly in AsciiDoc, for example. So I think context matters highly.”

Yves Hilpisch says “AsciiDoctor is my favorite these days. Clear syntax, nice output, fast rendering (HTML/PDF). Have custom Python scripts that convert @ProjectJupyter notebooks into text files from which I include code snippets automatically.” His scripts are in this GitHub repository.

Markdown

Robert Talbot recommends “Markdown in a plain text editor, with Pandoc on the back end for the finished product. This is assuming that the book is mostly text. If it involves code, I might lean more toward Jupyter and some kind of Binder based process.” Here’s a blog post Robert wrote on the topic.

One person recommended “writing Markdown then using pandoc to pass to LaTeX”, which is an interesting chain.

Visual Studio Code got a few mentions: “I haven’t written a full book using it, but VS Code plus markdown preview and other editing plugins is my current go-to for small articles”

Jason Moore recommends “your preferred text editor + RestructuredText + Sphinx = pdf/epub/html output; wrote my dissertation with it 6 years ago and was quite happy with the results.”

Matt Harrison uses his “own tools around rst (with conversion to LaTeX and epub).”

Raffaele Abate recommends “ScrivenerApp: I’ve used their Linux beta in past for a short, nonscientific, book and I can say it’s an amazing software for this purpose. I’ve read that is usable also for scientific publishing with profit.”

Lak Lakshmanan wrote, “I used Google docs for my previous book and for my current offer. Not as composable as latex, but amazing for collaboration. Books need to fine-grained reviews and edits by several people spread around the world. Nothing like Google docs for that.”

And the winner is…

For now I am working in LaTeX with Texmaker, but I have run it through pandoc to generate AsciiDoc, and that seems to work well. I will work on the book and the conversion process at the same time. At some point, I might switch over to editing in AsciiDoctor. I also need to do a test run with O’Reilly to see if they can ingest the AsciiDoc I generate.

A few weeks ago I posted “Lions and Tigers and Bears“, which poses a Bayesian problem related to the Dirichlet distribution. If you have not read it, you might want to start there.

Now here’s the follow-up question:

Suppose there are six species that might be in a zoo: lions and tigers and bears, and cats and rats and elephants. Every zoo has a subset of these species, and every subset is equally likely.

One day we visit a zoo and see 3 lions, 2 tigers, and one bear. Assuming that every animal in the zoo has an equal chance to be seen, what is the probability that the next animal we see is an elephant?

A recent paper in the New England Journal of Medicine, Attention Deficit–Hyperactivity Disorder and Month of School Enrollment, compares diagnosis rates for ADHD for children born in August and children born in September. In states that have a September 1 cutoff for children to start school, students born in September are generally the oldest in their class, and children born in August the youngest.
And it turns out that children born in August are about 30% more likely to be diagnosed with ADHD, plausibly due to age-related differences in behavior.
The analysis in the paper uses null-hypothesis significance tests (NHST) and focuses on the difference between August and September births. But if it is true that the difference in diagnosis rates is due to age differences, we should expect to see a “dose-response” curve with gradually increasing rates from September to August.
Fortunately, the article includes enough data for me to replicate and extend the analysis. Here is the figure from the paper showing the month-to-month comparisons.
Note: there is a typographical error in the table, explained in my notebook, below.
Comparing adjacent months, only one of the differences is statistically significant. But I think there are other ways to look at this data that might make the effect more apparent. The following figure, from my re-analysis, shows diagnosis rates as a function of the difference, in months, between a child’s birth date and the September 1 cutoff:
￼
For the first 9 months, from September to May, we see what we would expect if at least some of the excess diagnoses are due to age-related behavior differences. For each month of difference in age, we see an increase in the number of diagnoses.
This pattern breaks down for the last three months, June, July, and August. This might be explained by random variation, but it also might be due to parental intervention; if some parents hold back students born near the deadline, the observations for these months include some children who are relatively old for their grade and therefore less likely to be diagnosed.
We could test this hypothesis by checking the actual ages of these students when they started school, rather than just looking at their months of birth. I will see whether that additional data is available; in the meantime, I will proceed taking the data at face value.
I fit the data using a Bayesian logistic regression model, assuming a linear relationship between month of birth and the log-odds of diagnosis. The following figure shows the fitted models superimposed on the data.
Most of these regression lines fall within the credible intervals of the observed rates, so in that sense this model is not ruled out by the data. But it is clear that the lower rates in the last 3 months bring down the estimated slope, so we should probably consider the estimated effect size to be a lower bound on the true effect size.
To express this effect size in a way that’s easier to interpret, I used the posterior predictive distributions to estimate the difference in diagnosis rate for children born in September and August. The difference is 21 diagnoses per 10,000, with 95% credible interval (13, 30).
As a percentage of the baseline (71 diagnoses per 10,000), that’s an increase of 30%, with credible interval (18%, 42%).
However, if it turns out that the observed rates for June, July, and August are brought down by red-shirting, the effect could be substantially higher. Here’s what the model looks like if we exclude those months:
Of course, it is hazardous to exclude data points because they violate expectations, so this result should be treated with caution. But under this assumption, the difference in diagnosis rate would be 42 per 10,000. On a base rate of 67, that’s an increase of 62%.
Here is the notebook with the details of my analysis:

Suppose we visit a wild animal preserve where we know that the only animals are lions and tigers and bears, but we don’t know how many of each there are.

During the tour, we see 3 lions, 2 tigers, and one bear. Assuming that every animal had an equal chance to appear in our sample, estimate the prevalence of each species.

What is the probability that the next animal we see is a bear?

Now that we solved the appetizer, we are ready for the main course…

Suppose there are six species that might be in a zoo: lions and tigers and bears, and cats and rats and elephants. Every zoo has a subset of these species, and every subset is equally likely.

One day we visit a zoo and see 3 lions, 2 tigers, and one bear. Assuming that every animal in the zoo has an equal chance to be seen, what is the probability that the next animal we see is an elephant?