Dispatch views
These are the figures we read the open data with. They are built from one dispatch record and nothing else, drawn by a script we publish. The record itself is row level and is not published. Rebuilt every Monday, Wednesday and Friday.
Seven days across, twenty four hours deep, and the height of each block is how many transmissions that hour of that weekday carried on an average day it was observed. Drag the slider to turn it. Hover or tap a block to read it exactly.
The height is the only thing encoding the quantity. The blocks are one colour on purpose: a second channel saying the same thing would make the figure look more precise than the counting behind it.
The same peaks the figure shows, as text, because a three dimensional figure is not readable by everyone and the numbers should not depend on seeing it.
Our capture is a fraction of what actually gets dispatched, which is the first thing anyone should ask about it. So here is the hour of day profile of our file laid over the same profile from the Florida NFIRS record, which is a different and much larger record of the same state. If the two curves have the same shape, the slice is at least not arriving at odd hours.
Both are drawn as shares, because the sample sizes are four orders of magnitude apart and plotting counts would put one of them flat on the axis. Both denominators are in the caption.
Between ten at night and six in the morning, the single longest run with no dispatch on it. One mark per night, in order, for every night in the window. It is not averaged, and it is not going to be: the average of a night with one long run and a night cut into six pieces describes neither of them.
Nights where the receiver heard nothing at all that day are drawn as absent rather than as restful, since from inside the file those two look the same.
This is a count of dispatches in a window of the clock. It is not a sleep measurement, it says nothing about any individual, and no health claim is made or implied by it.
The nature of a call arrives as whatever the dispatcher said, so it is uneven text and not a code. Grouping it takes a judgement, and the number of distinct spellings feeding each group is printed beside the group so you can see how much judgement. The unclassified pile is drawn at its real size rather than dropped, because it is the largest single statement the field makes about itself.
Unit is the finest grain on this page. There is no station figure and no per person figure anywhere on it: pushing a department rate down to a station misstates the busy ones badly, and nothing about an individual is reachable from a radio transmission or belongs on a public page.
Every figure above has a denominator, and some of those denominators are much smaller than the row count, because several fields are blank on most rows. Here they are, rather than left for someone to find.
Two things are absent by decision rather than by accident. There is no station level figure on this page, because a department rate pushed down to a station is a well documented error and on this data it misstates the busiest station several times over. And there is no score: the measures built on top of this activity are held until the review that gates them is done. What is drawn here is counting.