December 2, 2021

Internet use up

The Herald has numbers from Chorus on internet data use, which is up since last October. Their data is broken down by region. I noticed that Auckland was at the top and wondered how much of this was better internet access in Auckland and how much was just larger households. Here’s a graph (click to embiggen). I had to guess that the ‘Hamilton’ region meant Waikato, and the table is missing Marlborough. Also, my data source for householdĀ  size had separate figures for Nelson and Tasman, but it should be basically right.

That’s actually more of an impact of household size than I expected. Also, I was a bit surprised that the West Coast is above the fitted line, saying that it has more internet use than you’d expect from household size, but I suppose that’s what you’d hope when people are spread out a lot.

The regression line is a bit unreliable with such a small dataset, and leaving out Auckland weakens the evidence for a relationship quite a bit (though it doesn’t actually change the fitted line very much). It’s worth thinking about alternative explanations. It’s reasonable that internet use would scale with household size (and I did think of this before looking at the data), but it could also be that Auckland has larger household size and more internet use because it’s a city

avatar

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »

Comments

  • avatar
    Megan Pledger

    I suspect big households have lower average age and, that age is also associated with internet use.

    2 years ago

    • avatar
      Thomas Lumley

      That’s probably true *within* region, but Auckland doesn’t have a much lower average age than the rest of the country.

      2 years ago

      • avatar
        Megan Pledger

        Here is a toy example of what I mean.
        Suppose in region 1 there are 3 households with ages
        h1: 50 50 20 15 14
        h2: 65 60 57
        h3: 75 74
        and in region 2 there are three households with ages
        h1: 50 50 20 16
        h2: 65 60
        h3: 75

        The average age in each region is 48 but the average household age is younger in region 1. If internet usuage lessens with increasing age then it looks like bigger households have more internet usuage per person.

        2 years ago

    • avatar
      Steve Curtis

      Auckland rents might mean theres more shared households as well to boost the averages
      However the straight line seems to fit our 2 person household very nicely down there , maybe 150GB off the graph in the bottom left

      2 years ago

  • avatar
    Nick Iversen

    At what household size would the internet usage be zero? Looks like about 1.6. I could ask – does that mean single person households don’t use the internet? But I know that’s the wrong question. The right question to ask is – are there any regions in NZ that have a household size of 1.6 and what is their internet usage?

    2 years ago

    • avatar
      Thomas Lumley

      The internet data aren’t reported by household size, but I don’t think extrapolating that far outside the range of the data makes a lot of sense anyway.

      2 years ago

      • avatar
        Thomas Lumley

        That is, I don’t see why the relationship should necessarily be linear over that sort of range, and it’s not possible to tell from these data

        2 years ago

  • avatar
    Jim Vine

    Is there a way of visualising this that would help to illustrate what component might be down to household size and what part remains unexplained by that?

    Eyeballing the line, I see about 300GB at 2.3 people. Scaling that up to 3 people would give (300/2.3*3) ~= 390GB. So that means the 600GB seen in Auckland well above the level.

    You could normalise the data usage to mean GB per household member, I guess. But then I suppose the question would be what to plot it against.

    2 years ago

    • avatar
      Thomas Lumley

      I’m not sure it’s worth doing that with these data — the range of the observed points is small enough that I’d be reluctant to force a straight line with this cluster of points and (0,0), which is what it would amount to. Almost none of the information would come from observable variation; it would almost all come from the straight line between (0,0) and the average of these points.

      I think you’d need individual household data to make that worthwhile.

      2 years ago