Is there a citytemp.dta version with the city names?

[Robert Grant]

I like using the -sysuse citytemp- dataset for teaching visualisation and I've often speculated about what might lead particular dots to be unusual... anybody know if there is a version out there with city names?

[Nick Cox]

I have never heard of such a version, but let's try some geography such as I did study once:

1. continental interiors have greater range than coastal regions

2. mountain areas often show greater range than lowland areas

#1 and #2 can reinforce each other.

Northern California and southern Oregon have remarkably low annual range, a story of ocean currents and so forth as well as #1 and #2.

[Robert Grant]

When I first show a scatterplot like this, many people are bamboozled. Then we start to unpick it. Mountain region has a steeper "slope" so that means greater range of summer temperatures -- maybe altitude effect? But it would be nice to show that this is true by referring to the data. Likewise a couple of outliers in Pacific region (which is understandable if that goes from San Diego to Anchorage, let alone maybe Hawaii -- but I don't know what's in and what's not

).

[Leonardo Guizzetti]

I suspect this is data scraped from Wikipedia or the USGS. You might try either of those to build at least a highly similar dataset.

[Nick Cox]

My understanding is that the dataset comes from US Census reports. Supposedly they carry all kinds of auxiliary data beyond those collected from people every decade (or so — 2020 census was delayed). It predates web scraping!

[Leonardo Guizzetti]

Hmm that’s interesting, Nick. I didn’t know that about the US Census.