free range statistics

Shiny can be an effective platform to quickly build data-intensive web applications that otherwise would not be commercially viable. The rationale for using Shiny at the right time is convenience, cost, and statistical and graphics power.

An animated map of UK Royal Navy ship locations during World War I.

I reproduce the analysis of data from a recently published experiment on the impact on Australians' and New Zealanders' attitudes to overseas aid of being exposed to writing about Chinese aid in the Pacific. Along the way I muse about the Table 2 fallacy, and try to avoid it while still using multiple imputation, bootstrap and adjusting for covariates to slightly improve the original analysis.

I play around with Hamlet's text to set it up for easy data analysis. Hamlet is awesome, this post is really just an excuse to spend time with it; but it does perhaps start to put together something useful about data models for text.

Two huge surveys of Facebook users seem to provide valuable new information on how the world is responding to Covid-19, but I am very unsure about whether they have potential to enable earlier detection of outbreaks.

I try out biterm topic modelling on a free text question in the 2017 New Zealand Election Study about the most important issue in the election.

I have a go at using the Processing programming language, which was developed to help graphic designers but has a broad bunch of potential applications. While Processing proper sits on top of Java, there's a handy implementation in JavaScript.

I look at some unusual data where the median was higher than the mode, and show how to model it in Stan as a mixture of two negative binomial distributions.

I confront past nowcasts of effective reproduction number for Covid-19 in Victoria with the best hindsight estimate, and confirm that the nowcasts lag change in the 7-14 days leading up to the time they are made.

I have a go at synthesising data to re-create a controversial and much-criticised chart that used ordinary least squares to fit a line relating university subjects' costs per student to the number of students in each subject.