iiinova — where your digital self keeps evolving

US diesel has been retailing at above $6 per gallon for a while now and looks likely to stay there for a while longer. But still cheap compared to elsewhere. Only briefly was it more expensive in the USA than New Zealand (not counting the NZ road user charges). freerangestats.info/fuel-crisis/...

line chart showing USA and NZ retail petrol (gaoline) and diesel prices. The last few weeks NZ prices have been more than USA for diesel, but there was a time in September US prices were more. Gasoline is always much more expensive in NZ (because of taxes).

110

Peter Ellis @freerangestats.info · 23h

Great keynote today by Neil Pearce for the Australasian Epidemiological Association using asthma and kidney disease as egs. Argues much that is called 'causal inference' is really causal estimation, and inference is more like multi-source, multi-method detective work eliminating suspects one by one.

Lecturer in front of slides showing news articles claiming kidney disease in Sri Lanka is 'caused by pesticides', which turned out to be wrong, and action based on the hypothesis was economically and socially very costly.

070

Peter Ellis @freerangestats.info · 06/10/2026

It's difficult to find estimates of total world crude oil stocks. I did the best I could with the 38 countries that have consistently complete data in JODI. Key message - they're low, but not at crisis levels for a while. #rstats blog at freerangestats.info/blog/2026/10...

time series plot from around 2002 to present showing crude oil stocks of consistnetly reporting countries worldwide. currently at a low level of about 1.7b barrels, 20 days refinery cover. Projected to be down to 10 days of cover in 92 weeks.

010

Peter Ellis @freerangestats.info · 02/10/2026

I am looking forward to contributing to this event by talking about the challenges and opportunities of small area estimation methods, based on experience to date with real data in Pacific Island countries. This is an online workshop on 5 November, registration details in the attached image.

Promotional poster for the “Small Area Estimation in the Pacific: Current Challenges and Future Directions” online workshop. The workshop will be held on 5 November 2026 and brings together researchers, statisticians, practitioners and policymakers to discuss methodological advances, Pacific-relevant applications, data challenges, collaboration, and future capacity building in small area estimation. Registration is free and closes on 29 October 2026; the poster includes a QR code, registration link, and a list of confirmed and invited speakers from SPC, Australian universities and statistical agencies.

063

Peter Ellis @freerangestats.info · 20/09/2026

A small #rstats blog post following up on last week's on sample size and the central limit theorem. How well does BCa bootstrap do in comparison to older methods for a confidence interval for the mean from a skewed population? Better, but not a magic fix. freerangestats.info/blog/2026/09...

Scatter plot comparing coverage of 95% confidence interval of BCa bootstrap with traditional CI calculated based on asymptotic normality. For various sample sizes and population distributions. BCa does better, but still less than the 95% one would hope.

081

Peter Ellis @freerangestats.info · 18/09/2026

This is what I see, for example.

Claude asked about college attainment in US and UK, quotes back results from US Census Bureau and UK ONS.

000

Peter Ellis @freerangestats.info · 16/09/2026

But I don't think it makes as much difference to this exercise as people might think. In fact in my first goes I *was* using confidence intervals based on t (see attched) and it doesn't look a million miles different.

210

Peter Ellis @freerangestats.info · 16/09/2026

Inspired by a @f2harrell.bsky.social comment, I did some #rstats simulations about how big a sample needs to be for the central limit theorem to kick in and the sample mean be normally distributed for inference purposes. Yes, with skewed data the answer is "lots". freerangestats.info/blog/2026/09...

Scatter plot with 8 facets, each showing a different distribution, and the actual coverage of a 95% confidence interval for the mean at different sample sizes. Most values are well below 95%.

33315

Peter Ellis @freerangestats.info · 16/09/2026

What do we call this phenomenon, where we are assailed with the near-identical one line summary of something multiple times. Not to pick on Reuters at all, this is a much bigger systemic thing. Sometimes we can go through 5 or 6 layers of Russian dolls of the same sentence over and over again.

060

Peter Ellis @freerangestats.info · 28/08/2026

I see Samoa's vaccination and measles record are in the news. Coincidentally yesterday I gave a talk with these relevant charts. Nasty spike in DALYs lost from measles along with the outlier low point vaccinations.

000

Peter Ellis @freerangestats.info · 12/08/2026

US crude oil stocks recovered slightly in the latest weekly EIA data - little fishhook in the blue line in this chart. freerangestats.info/fuel-crisis/...

Line chart of US total crude oil stocks 2000 to present

010

Peter Ellis @freerangestats.info · 08/08/2026

I have added to my little personal collection of fuel-crisis-monitoring charts a plot of the retail prices of petrol/gasoline and diesel, in USA and New Zealand (where filling your car, quite rightly, costs US$7 per gallon). Blog and #rstats code at freerangestats.info/blog/2026/08...

Time series chart of petrol and diesel prices in New Zealand and USA since 2004, with the GFC, '$100 crude plateau', shale oil hitting the market, Russian invasion of Ukraine, and Iran attack on Iraq all marked.

150

Peter Ellis @freerangestats.info · 05/08/2026

Finally, of more limited interest perhaps, inflation in the Pacific showed signs of moderating, with prices in Fiji declining somewhat in July. Overall inflation in Pacific Islands has seen a very mixed impact from the Iran war.

time series plot of inflation (CPI) index in Tonga, Samoa, Fiji, New Caledonia, French Polynesia and Solomon Islands (listed from highest inflation in the Iran War period to the lowest)

030

Peter Ellis @freerangestats.info · 05/08/2026

Stocks in the key storage and distribution facility at Cushing, Oklahoma, increased above the 20 million barrels danger threshold.

time series plot of crude at Cushing, showing an increase in the last week above 20m barrels.

110

Peter Ellis @freerangestats.info · 05/08/2026

US crude oil stocks (including Strategic Petroleum Reserve) arrested their downwards slide in the latest data. They are now 20 weeks of linear decline from reaching a theoretical stress point of 30 days refinery cover; an improvement from 19 weeks a week ago. freerangestats.info/fuel-crisis/...

time series plot of US total crude oil stocks 2000 to now.

121

Peter Ellis @freerangestats.info · 01/08/2026

I couldn't find the exact annotated plot of total US oil stocks I wanted, so I had to make my own. At the rate of decline since its recent peak, there are 19 weeks left to having 30 days cover of refinery needs. Blog with #rstats code at freerangestats.info/blog/2026/08...

time series chart of total US crude oil stocks, including SPR, since 2000. Shows sharp decline in last few months. Point is labelled 712 million barrels, 41 days of refinery cover. Subtitle says that at rate of net decline since 3 April (10.4 million barrels per week), there are 19 weeks to a point of having only 30 days of refinery cover.

0133

Peter Ellis @freerangestats.info · 23/07/2026

Oil at the storage and distribution hub in Cushing, Oklahoma has gone down below 20 million barrels again.

time series of crude oil at Cushing, showing a decrease this week down to 19.37m barrels.

011

Peter Ellis @freerangestats.info · 08/07/2026

Very short #rstats blog post on polishing this chart of crude oil levels at Cushing, Oklahoma. This is one of the indicators I have been monitoring during the ongoing global fuel situation. Below 20m barrels is "tank bottoms" that might cause problems for markets. freerangestats.info/blog/2026/07...

time series chart of crude oil at Cushing in 2026. It goes up to above 30million barrels in April, down below 20 million for the last three weeks of June / July.

191

Peter Ellis @freerangestats.info · 01/07/2026

Yesterday's data show the levels of crude oil in storage at Cushing in Oklahoma have taken a turn upwards in the last week, showing there is some reality behind the talks/hopes of increased availability of oil.

Line chart of oil at Cushing, Oklahoma, showing it has gone about about 700k barrels in the last week to 19.7m, only just below the widely cited minimum operative level of 20m barrels.

000

Peter Ellis @freerangestats.info · 28/06/2026

I #rstats blogged about using the long term real GDP per capita estimates in the Maddison Project database, to draw charts comparing a country's "average" growth since 1990 to the actual trajectory. freerangestats.info/blog/2026/06...

1114

Peter Ellis @freerangestats.info · 25/06/2026

I don't really know if this is as big a problem as it seems, but recently people were saying that if Cushing oil stocks went below 20 million barrels it starts to freeze up the US oil market. Well, it is below there this week, and surely has a bit further down to go.

Time series chart of Cushing oil stocks showing down below 19m barrels this week

230

Peter Ellis @freerangestats.info · 23/06/2026

Discovered something very weird happening with the software I use to convert SVGs (my main image type on my blog) to PNGs to fall back on. It's like it's going through bad OCR - Liz Truss becomes Miz Trutt, George Canning has become Geoshe Cbooloh. And what has happened to that subtitle?

scatter plot of UK PM durations but with mangled names for some reason.

120

Peter Ellis @freerangestats.info · 23/06/2026

While certainly there has been a lot of churn in UK prime ministers in the last decade or so, it's not that remarkable in historical terms, or an obvious trend. In fact, the long terms of Thatcher and Blair stand out as much as the more recent burst of short terms.

Scatter plot of duration of PMs' terms by starting date of the term, showing a flat trend line. Long and short duration terms are labelled eg Liz Truss, Tony Blair, Margaret Thatcher, Robert Walpole.

chart of the rolling number of UK prime ministers in the last decade, which ranges from 1 to 9. Peak period was in the 1830s.

1112

Peter Ellis @freerangestats.info · 23/06/2026

Like many people I am interested in the rate of UK prime ministerial churn, so I wrote an #rstats blog on it and experimenting with various visualisation approaches. I quite like this Gantt-like chart showing the terms of all PMs since 1721. freerangestats.info/blog/2026/06...

Gantt chart style presentation of the terms of all UK prime ministers from Robert Walpole to Keir Starmer

2318

Peter Ellis @freerangestats.info · 19/06/2026

Interesting that when I search 'in the weights' with my own name I am only the 8th most prominent Peter Ellis, with a strength of 68; but when I search as freerangestats it still returns my personal name but now with a strength of 316.

220

Peter Ellis @freerangestats.info · 14/06/2026

Great history question from the ABC.

Question 1: Jurassic World Rebirth premiered last year in which city?

130

Peter Ellis @freerangestats.info · 14/06/2026

Blog post on seasonal adjustment in #rstats, using Samoan visitor arrivals since 2017 as a test case. Using seasonal, fable and feasts packages turns out to be easy. freerangestats.info/blog/2026/06...

time series plot of Samoan visitor arrivals, with a dramatic drop during the time of covid then recovery but not much growth since.

070

Peter Ellis @freerangestats.info · 06/06/2026

RStudio by @posit.co has taken a step backwards by my preferences. Nothing I can do in the options seems to stop Post Assistant pane jumping up and taking up real estate; and the "project" drop down box has vanished. Not a happy camper. Will probably have to look for an old version to install.

screenshot of RStudio showing a "Posit Assistant Not Enabled" pane popping up despite settings for Assistant all showing 'None' (this is after restarting RStudio); and no project drop-down list on the right like there used to be. All this was better a year ago.

670

Peter Ellis @freerangestats.info · 30/05/2026

I got this interesting variant

Claude claims Friday contains fish because fry is a word for small fish.

020

Peter Ellis @freerangestats.info · 13/05/2026

Comparing it to the duration of past wars (which range from 4 days in length to 10+ years), the US-Israel-Iran war is now longer than the Boxer Rebellion war, but not quite as long as the Falklands or the War for Kosovo (names used are those in the Correlates of War data).

Scatter plot with starting date of war on x axis, duration on y, size of points is mapped to number of battle deaths. Inter-state wars 1823-2003.

010

Peter Ellis @freerangestats.info · 13/05/2026

Of wars that get to the 74 day mark, the median total duration is 261 days. Which is grim news those who have noted that the Iran war will cause major economic problems if it lasts even another month. #rstats blog post, with @correlatesofwar.bsky.social data, at freerangestats.info/blog/2026/05...

empirical cumulative distribution chart of the duration of wars from 1823 to 2003, with the current Iran war marked at 74 days (about 0.33 of the cumulative distribution) and the median expectation duration of 261 days, conditional on at least 74 days.

1154

Peter Ellis @freerangestats.info · 24/04/2026

I like to think that 'the happy isles - you hit me, we hit you' says something about the duality of something or other.

License plate reads 'Solomon Islands - Hapi Isles. You hit me, we hit you.'

040

Peter Ellis @freerangestats.info · 08/04/2026

It's in the Wall Street Journal including Trump saying its a workable basis.

Same 10 point plan as above, in Wall St Journal.

010

Peter Ellis @freerangestats.info · 04/04/2026

But fuel oil in singapore has nearly doubled in price. I think what is going on is a combination of lag, competition in shipping, and fuel surcharges not being included in the container price index.

110

Peter Ellis @freerangestats.info · 04/04/2026

This is the short list of public, frequently updated prices I'm monitoring for the economic impact of the war in Iran on Asia-Pacific. Are there any suggestions or corrections #econsky? Why is the Intra-Asia Container Index not showing price increases comparable to the other indicators?

Table with 7 price indicators of the economic impact of the Iran crisis including Brent Crude, diesel in Singapore, jet fuel in Asia, intra-Asia merchandise freight, urea, fuel oil in Singapore, air fares. The growth in price since january 2026 ranges from 4% (intra-Asia container index) to 120% (jet fuel in Singapore).

141

Peter Ellis @freerangestats.info · 30/03/2026

Most Pacific island countries also do most of their household cooking with kerosene or gas fuel (and bear in mind that 'electricity' in this chart mostly means diesel, for now). 2/2

stacked area chart showing primary fuel used for cooking in 15 pacific island countries. For 2/3 of them it is mostly gas and kerosene.

061

Peter Ellis @freerangestats.info · 30/03/2026

Pacific island countries are *extremely* dependent on oil and gas products - particularly diesel - for electricity generation. Brief blog with #rstats code at freerangestats.info/blog/2026/03... 1/2

Stacked bar chart of Pacific island countries plus New Zealand and Australia showing share of electricity by source. For most countries the plot is a blaze of red meaning 'oil'.

1198

Peter Ellis @freerangestats.info · 22/03/2026

One of the consequences is extraordinarily high remittances rates. Anyway, this is all expanded on in the post at freerangestats.info/blog/2026/03....

120

Peter Ellis @freerangestats.info · 22/03/2026

Four out of ten of the biggest Pacific Islander cities in the world are in Australia, New Zealand and USA.

120

Peter Ellis @freerangestats.info · 22/03/2026

This all contributes to many Pacific countries and territories having more of 'their' people residing overseas (New Zealand, Australia and USA mostly) than in their origin country.

110

Peter Ellis @freerangestats.info · 22/03/2026

One of the best ways of classifying Pacific island countries and territories is into the 13 that have easy access to a big, richer country and the 9 that don't.

120

Peter Ellis @freerangestats.info · 22/03/2026

I have blogged at more length than usual on Pacific #demography issues, tying together with narrative the charts produced in six #rstats posts starting late last year. The theme is the obvious that people movement is critically important in the Pacific. freerangestats.info/blog/2026/03...

1101

Peter Ellis @freerangestats.info · 16/03/2026

Buried about 120m apparently. Its been studied! Claude recommends 100-150kt devices about 200m apart.

120

Peter Ellis @freerangestats.info · 09/03/2026

Brent Crude has nearly reached the 2022 peak.

Time series line chart of price of Brent Crude oil from May 2021 to March 2026, showing current price at $117 and rising almost vertically, and the peak in 2022 at about $128.

050

Peter Ellis @freerangestats.info · 08/03/2026

Not quite nothing but nearly. These are as % of GDP. But it's not impossible that PNG's stats aren't being collected as well as some of the other countries.

020

Peter Ellis @freerangestats.info · 08/03/2026

Remittances - payments from relatives or other contacts abroad - are astonishingly high in some Pacific island countries. I round out my series creating visuals related to Pacific #demography issues with a blog post with #rstats code at freerangestats.info/blog/2026/03...

Bar chart showing remittanes as a % of GDP, very high for Tonga (>40%), Samoa, Vanuatu, Marshall Islands and other Pacific countries, compared to much lower levels for low income countries gloablly, small states, etc.

150

Peter Ellis @freerangestats.info · 28/02/2026

Population pyramids can highlight radically different demographic trajectories; in this case of two neighbouring coral atoll nations, one of which has free people movement to a large, rich country, and one of which doesn't. Blog with #rstats code at freerangestats.info/blog/2026/03...

050