about free range statistics

free range statistics is the personal blog of Peter Ellis.

I'm an Australasian professional statistician and data scientist with a background running analytics/stats/data science teams and multi million dollar programs in the public sector. I am the Director of the Statistics for Development Division at the Pacific Community.

Much of my career has been in management and evaluation of large overseas aid programmes, but since 2011 I have been working on a range of economic data issues, and on data and statistical capability to support diverse organisational goals. My current role combines the two halves of my career - as an international development professional, and a data and statistics expert and manager - into one.

My interests include economics, econometrics, complex surveys, time series, spatial statistics, text mining, R, data visualisation, data warehousing, data governance, and anything that involves putting data and analysis in the hands of as broad a range of people as possible.

The views expressed in this blog are very definitely my own, not those of my organisation or any government. I might be blogging about data and analysis issues that crop up in work, but anything controversial that relates to my work has crept in by mistake and will be removed if and when I notice it.

• ggseas package which allows you to easily incorporate seasonal adjustment and seasonal decomposition into an exploratory data analysis workflow based around ggplot2.

• Tcomp package which provides data from the 2010 Tourism Forecasting Competition described in Athanasopoulos et al.

• nzelect package which provides conveniently tidied up election results for New Zealand; so far limited to aggregate results from the 2014 general election.

• nzcensus package which provides a range of calculated demographic variables at meshblock, area unit, territorial authority, and regional council level from the New Zealand census. This package was originally part of the nzelect project but can now be installed and used separately. The source is the Statistics New Zealand census meshblock data. The package is too large to publish on CRAN so is only available from GitHub.

I also host on my GitHub page the forecastHybrid package, which makes it easy to estimate ensemble or hybrid forecasts using a range of the methods in Rob Hyndman's forecast package. I'm a contributor to that package but David Shaub is the primary author and maintainer.