free range statistics

I expand on my last post, to see if the relationship between depression and voting for Trump at county-level persists when you control for the racial composition of counties (it doesn't).

Multi-level modelling with spatial auto-correlation! I look at county level data on incidence of depression in 2020, and voting for Trump in the 2024 US Presidential election, and conclude that there's something there, but of course there are lots of potential explanations of what is behind the relationship.

I explore death rates by cause of death with OECD data, for the USA and other countries. Causes of death that are relatively high in the USA include assaults, accidents, suicides; diseases of the nervous systems (including Alzheimer's); and diseases of the circulatory system (including heart attacks).

I write a function to simulate Ponzi schemes with various types of 'investor' growth, withdrawal rates, and extraction by the scammer / owner of the scheme.

I look at the two different sorts of design effects that Stata will report for estimates from sub-populations of a complex survey, which vary depending on whether or not the hypothetical simple random sample we are comparing the complex survey to has the same sub-population sample sizes as the actual sample.

I compare some different ways of forcing the coefficients in a regression to form a simplex, all greater than zero and adding up to exactly one. Two methods - quadratic programming, and explicit modelling of the coefficients from a Dirichlet distribution - give essentially identical results that match the data generating process well.

I was honoured to give the third and final Ihaka Lecture in the 2024 series. My talk had the theme "Making R work in government".

I explore the number of ways to make a prime number as the sum of squares of three positive integers.

Stepwise variable selection is bad and dangerous, and you shouldn't do it. It increases false positives. It drops variables that should be in the model. It gives biased estimates for regression coefficients. The problems are worse for smaller samples; higher correlation between the X variables; and models with weaker explanatory power for the y (i.e. lower R-squared).

I familiarise myself with the Australian Bureau of Statistics' statistical Standard on sex and gender, and play around with some data from the Australian General Social Survey that has outputs reported by persons' sexual orientation.