Poetry is just the evidence of life. If your life is burning well, poetry is just the ash
•
• burn something

Math geek? If you like the clean geometric design of the type posters, you may enjoy something even more mathematical. Design that transcends repetition: Art of Pi, Phi and e posters.

Sometimes to understand the whole, we need to look more closely at its parts. Completely safe for work.

I include Emigre's Platelet because it's such a goofy and fun font. One look at the lower case *b* and you know this isn't a type face that wears a tie.

To me the ampersand is the letter I always say yes to.

Below, sixteen ampersands from different fonts are centered vertically and horizontally. Look how they blend together to reveal a consensus letter.

Finding inspiration in this IKEA Olunda poster of Akzidenz-Grotesk, I've rendered the full alphabet in sixteen faces. Look at how agreeable the T's are. But the J's — oh, the J's — a riot.

Building on last month's column about Bayes' Theorem, we introduce Bayesian inference and contrast it to frequentist inference.

Given a hypothesis and a model, the frequentist calculates the probability of different data generated by the model, *P*(data|model). When this probability to obtain the observed data from the model is small (e.g. `alpha` = 0.05), the frequentist rejects the hypothesis.

In contrast, the Bayesian makes direct probability statements about the model by calculating P(model|data). In other words, given the observed data, the probability that the model is correct. With this approach it is possible to relate the probability of different models to identify one that is most compatible with the data.

The Bayesian approach is actually more intuitive. From the frequentist point of view, the probability used to assess the veracity of a hypothesis, P(data|model), commonly referred to as the *P* value, does not help us determine the probability that the model is correct. In fact, the *P* value is commonly misinterpreted as the probability that the hypothesis is right. This is the so-called "prosecutor's fallacy", which confuses the two conditional probabilities *P*(data|model) for *P*(model|data). It is the latter quantity that is more directly useful and calculated by the Bayesian.

Puga, J.L, Krzywinski, M. & Altman, N. (2015) Points of Significance: Bayes' Theorem *Nature Methods* **12**:277-278.

Puga, J.L, Krzywinski, M. & Altman, N. (2015) Points of Significance: Bayes' Theorem *Nature Methods* **12**:277-278.

In our first column on Bayesian statistics, we introduce conditional probabilities and Bayes' theorem

*P*(B|A) = *P*(A|B) × *P*(B) / *P*(A)

This relationship between conditional probabilities *P*(B|A) and *P*(A|B) is central in Bayesian statistics. We illustrate how Bayes' theorem can be used to quickly calculate useful probabilities that are more difficult to conceptualize within a frequentist framework.

Using Bayes' theorem, we can incorporate our beliefs and prior experience about a system and update it when data are collected.

Puga, J.L, Krzywinski, M. & Altman, N. (2015) Points of Significance: Bayes' Theorem *Nature Methods* **12**:277-278.

Oldford, R.W. & Cherry, W.H. Picturing probability: the poverty of Venn diagrams, the richness of eikosograms. (University of Waterloo, 2006)

Celebrate `pi` Day (March 14th) with splitting its digit endlessly. This year I use a treemap approach to encode the digits in the style of Piet Mondrian.

The art has been featured in Ana Swanson's Wonkblog article at the Washington Post—10 Stunning Images Show The Beauty Hidden in `pi`.

I also have art from 2013 `pi` Day and 2014 `pi` Day.

The split plot design originated in agriculture, where applying some factors on a small scale is more difficult than others. For example, it's harder to cost-effectively irrigate a small piece of land than a large one. These differences are also present in biological experiments. For example, temperature and housing conditions are easier to vary for groups of animals than for individuals.

The split plot design is an expansion on the concept of blocking—all split plot designs include at least one randomized complete block design. The split plot design is also useful for cases where one wants to increase the sensitivity in one factor (sub-plot) more than another (whole plot).

Altman, N. & Krzywinski, M. (2015) Points of Significance: Split Plot Design *Nature Methods* **12**:165-166.

1. Krzywinski, M. & Altman, N. (2014) Points of Significance: Designing Comparative Experiments *Nature Methods* **11**:597-598.

2. Krzywinski, M. & Altman, N. (2014) Points of Significance: Analysis of variance (ANOVA) and blocking *Nature Methods* **11**:699-700.

3. Blainey, P., Krzywinski, M. & Altman, N. (2014) Points of Significance: Replication *Nature Methods* **11**:879-880.

In an audience of 8 men and 8 women, chances are 50% that at least one has some degree of color blindness^{1}. When encoding information or designing content, use colors that is color-blind safe.