The longest and the shortest

The two extremes

The long one is Home E, a Thursday, at the place that home had set up for work. It began at half past eight and ended at twenty to twelve, and it ended with a dash: the work finished, nothing in the room moved anybody.

The short one is Home G, a Monday, at a window seat. It began at ten past nine and ended at sixteen minutes past, with the word elsewhere. Something was in another room and the person went to get it and did not come back to the window. The practical question is whether a change improves completed work without creating unnecessary surveillance, which is why reviewers may look at the current implementation details before choosing a baseline and access policy.

Neither is an error and neither was checked more carefully than any other entry, which is worth saying because extremes attract scrutiny that the middle never gets, and a record that verifies only its outliers will end up with suspiciously clean outliers. Readers repeating the exercise can place this observation beside material from King's College London.

The shape

The distribution has a single hump around three quarters of an hour and a long tail to the right. It is not bimodal. There is no cluster of very short sittings sitting apart from the rest, which is what a record dominated by interruption would look like.

Ten per cent of sittings ran under twenty minutes. Those twenty-two are worth their own paragraph, because a tenth is small enough to ignore and large enough to matter.

The twenty-two short ones

Fourteen of the twenty-two ended with one of the nine words. Eight ended with a dash.

That eight is the interesting part. A sitting of under twenty minutes that ended because the work ended is somebody sitting down to do one small thing and then getting up again, and nothing about it is a failure of anywhere. The record contains more of those than a reader primed to look for interruption would guess.

Of the fourteen, the commonest endings were elsewhere (five) and wanted (four) — which is to say, short sittings are ended by somebody or something needing to be somewhere, rather than by the room becoming unpleasant.

Light and temperature almost never end a short sitting. They take time to change.

The endings that depend on the room sort themselves to the long sittings, and the endings that depend on other people sort themselves to the short ones. That pattern holds across all nine homes.

The sixteen long ones

Sixteen sittings ran two hours or more. Thirteen of them ended with a dash.

Three did not: one temperature, one wanted, one light. All three happened in the afternoon, and in all three cases the thing that ended them had been building for the whole sitting.

The thirteen dashes are the quiet centre of this record. A two-hour stretch that ends because the work ends is a dwelling doing exactly what is asked of it, and there are thirteen of them in forty-five days, spread across six of the nine homes.

What the tail is not

It is tempting to read the right tail as a quality measure — long sittings are good, short ones are bad — and the record does not support that in either direction.

Home D has the longest median in the study, seventy-three minutes, and two places, one of which is a bedroom. Home F has the shortest, thirty- five, and six places, a garden-facing back room and the most varied day in the record.

Nothing here says which of those is the better week. The log asked how long people stayed, not whether staying was preferable.

Why the median and not the mean

The mean sitting is 48.8 minutes and the median is 48. They are almost identical, which in a distribution with a tail this long is slightly surprising and happens because the tail is thin.

The median is used throughout anyway, for a reason that holds whether or not it matters in a particular case: a mean on two hundred and twenty-five observations can be moved several minutes by one long entry, and this record has entries of a hundred and ninety-four minutes in it. A figure that one line can shift is a figure that invites checking the line rather than the finding.

The entry that was nearly excluded

The hundred-and-ninety-four-minute sitting was queried, because a single unbroken stretch of more than three hours is unusual enough to suggest a household had logged two sittings as one.

It had not. Asked directly, the household described the morning in enough detail to be confident, including the fact that the person had stood up twice without leaving the place — which, under the rule set out in the definition, is still one sitting.

It stays in. But it is the clearest illustration of what the unit cannot see: three and a quarter hours in this record and two minutes on one's feet in the middle of it look exactly the same.