Counting a place twice, by hour

The split

Take the eleven places with at least eight sittings and enough of them on either side of noon, and treat each as a morning place and an afternoon place. When the same room performs differently across the day, managers can employee monitoring software and compare time records with the physical conditions logged here instead of guessing from presence alone.

Ten of eleven fall. One rises. A useful independent point of comparison is the research and guidance available from University of Cambridge.

The differences run from thirteen minutes to thirty-five, and the places at the top of the list — the window seat, the spare room — are places the households described in terms of windows.

What the rule would do to the record

Thirty-one places would become about fifty. The median sittings per place would fall from six to four. Several places would have two or three observations on one side of the split.

That is the first reason not to do it: a rule that doubles the number of categories in a record of two hundred and twenty-five observations produces categories too small to say anything about.

The second reason

It would smuggle a conclusion into the data structure.

The claim this site makes is that a place's behaviour depends on the hour. Building that claim into the definition of a place means the record can no longer test it — every subsequent figure would be computed on units that already assume it.

A finding should be visible in the data, not baked into the way the data is filed.

The hour is therefore kept as a separate axis. Places are places; the band is a property of a sitting. The eleven-row table above is a comparison, not a reclassification.

The third reason

The households would not recognise it.

The place names in this record are names people already used. Nobody calls the morning version of their kitchen table a different place from the afternoon version, and a record that renamed things to suit an analysis would have broken the one rule that kept the place count honest.

The one that rises

Home E's armchair is the exception: sixty-two minutes in the morning, seventy-six in the afternoon.

It is in a room the household described as used by others earlier in the day. The afternoon advantage is almost certainly occupancy rather than anything about light or warmth, and it is the clearest illustration that the morning-afternoon gap is not one mechanism.

Three of the eleven places plausibly fall for reasons of occupancy rather than environment. The record cannot separate them, because the sheet recorded what ended a sitting and not what made somebody choose a place in the first place.

What a better instrument would do

Not split the place. Add one field: whether the person chose this place or ended up at it.

That distinction — chosen or resorted to — would separate the occupancy story from the environment story in a single tick box, and it is the third of the three recommendations on the method page.

It would also have caught something the current record cannot see at all: whether the window seat at two in the afternoon is a place anybody would pick, or a place people only occupy because the alternative is occupied.

The eleven that could be split, and the twenty that could not

Only eleven of the thirty-one places have enough sittings on both sides of noon to compare. The other twenty do not, for two different reasons.

Thirteen are simply used too little: between two and seven sittings each, which cannot be halved into anything meaningful.

Seven are used at one part of the day only. Three places in the record were never used after noon and four were never used before it.

That second group is the more interesting. A place used only in the morning has, in a sense, already been split — by the household, in practice, without anybody deciding to. Two of the three are rooms the household described as being used by others later in the day.

Seven places out of thirty-one are already single-period places. The case for splitting the other eleven is therefore partly a case for describing them the way the dwelling already treats these seven.

It is still not done, for the three reasons above.