Analyze This! – A Public Dataset, Charted Live and Explained by the Statistics

Analyze This! is our ongoing series that takes an openly available dataset, loads it into QuantumLayers, and lets the charts and the statistics tell the story. Every figure in these posts is drawn live from a public dataset, with no spreadsheet, no export, and no hard-coded number anywhere in the page. Here is how each one is built.

What the Series Is

Analyze This! runs here on QuantumLayers, and each entry follows the same simple contract. We pick a public dataset large enough to hold a real pattern, connect it to QuantumLayers, and publish what the platform finds. Four are live so far, and although their subjects could hardly be further apart, the machinery behind them is identical.

The Datasets So Far

The earthquakes post takes every earthquake the U.S. Geological Survey recorded over a single 30-day window, 10,877 events in all, which works out to roughly one every four minutes. The overwhelming majority are far too faint to feel: about 78 percent fall below magnitude 2.3, and the median event is a barely perceptible 1.47. The day-to-day count is remarkably steady, holding between roughly 350 and 490 recorded quakes, while a lone magnitude 7.7 sits at the top of the range. Depth tells its own story. Half of all quakes originate within about 7.5 km of the surface, but a long tail of deep-focus events reaches down to 665 km, a lopsidedness QuantumLayers flags as real physics rather than bad data. Alaska and the U.S. West Coast, the most heavily instrumented regions, account for most of the catalogue, with Alaska alone reporting more than 2,000 events in the month.

The meteorites post loads NASA’s complete catalogue of known meteorite landings, 45,716 of them, with records reaching back to the year 860. For most of history the catalogue grew by a trickle, a handful of witnessed falls a year, then it jumps sharply after 1976 as organized expeditions began sweeping Antarctica and the world’s hot deserts, where dark meteorites stand out against pale ground. Only 1,107 of the tens of thousands on record were ever actually seen to fall; everything else was found long after landing. The rocks themselves are strikingly uniform, dominated by a few ordinary chondrite classes, with type L6 alone close to a quarter of everything classified. Mass is where the data turns extreme: the typical meteorite weighs about 48 grams, roughly a golf ball, yet the distribution runs all the way up to the 60 tonne Hoba meteorite in Namibia, one of the most lopsided spreads the series has charted.

The volcanoes post takes the Smithsonian’s catalogue of 958 Holocene volcanoes and asks a human question of it: who lives nearby. The answer is more than 2.6 billion people within 100 km of one. That crowding is not evenly spread. A short list of countries, led by the United States, Indonesia and Japan, holds most of the catalogue, and the single most crowded volcano, Java’s Gede-Pangrango, sits within reach of more than 40 million people. Elevation barely predicts any of this: the correlation between a volcano’s height and its nearby population is only about 0.16, so what fills the ground around a volcano is human geography, not altitude.

The squirrels post is the lightest subject and, in some ways, the most surprising demonstration. It loads the 2018 Central Park Squirrel Census, 3,023 individual sightings hand-recorded by volunteers, and finds real structure in it: about 83 percent of squirrels are gray, roughly 72 percent were seen at ground level, and the platform’s automated tests surface a clean split between climbing and foraging as two mutually exclusive modes of squirrel behavior. Plotting every sighting by its coordinates even redraws the outline of the park itself, entirely from where the squirrels happened to be standing.

From Public Dataset to Typed Columns

There is no spreadsheet stage, and nothing is downloaded by hand. The source is connected directly as a dataset inside QuantumLayers, which ingests every column and assigns it a type on the way in. The earthquake feed arrives as 22 columns of magnitudes, depths, coordinates, timestamps, and reporting networks, and QuantumLayers reads all of them without anyone reshaping the file first. That single step, connect and type, is the whole preparation phase.

Why Every Chart Is Drawn Live

The charts in an Analyze This! post are not screenshots. When the page loads, the reader’s browser makes one POST to QuantumLayers’ public chart endpoint, ql_get_chart_data, and gets back a ready-to-draw Chart.js configuration, which it then paints with Chart.js. Because the dataset is marked public, that request needs no API key, no login, and no server of ours in the middle. Nothing in the article is a fixed figure typed into the HTML. Each chart reflects whatever the dataset holds at the moment you open the page, so the post stays current on its own as the underlying data is revised.

One Chart, Two Front Doors

Every figure ships with a second way in. Alongside the embedded chart sits a link to a standalone ql-chart page hosted on QuantumLayers, and that page rebuilds the same visualization from nothing but its URL. The dataset, the chart type, and the columns are all encoded as parameters, so a histogram of earthquake magnitudes or a doughnut of meteorite classes is a plain link anyone can open, share, or point their own page at. The posts use histograms, time series, doughnuts, scatter plots, and box plots, each one both embedded in the article and reachable as its own URL.

Where the Findings Come From

Charts show the shape, and the statistics pin it down. For each dataset we run QuantumLayers’ automated statistical analysis across every numeric column, which tests distributions, flags outliers, and measures how strongly one field moves with another. That is where the parts of these posts that read like conclusions come from. In the meteorite catalogue, the masses form one of the most lopsided distributions we have charted, with a skewness near 57, because that single 60 tonne outlier towers over a typical rock of about 48 grams. In the earthquake set, an analysis of variance shows that how a quake is measured is tied to where it is, which is methodology rather than geology. The platform ranks these findings by importance and writes them in plain language, so the post inherits a result that has already been validated instead of a chart left for the reader to interpret.

Why We Build Them This Way

Each post is also a working demonstration. The point we make elsewhere, that the useful output of an analytics tool is the finding and not the chart, is easier to show than to argue, so Analyze This! shows it on data anyone can inspect. Because everything runs on public datasets and a public endpoint, there is nothing to take on trust. The numbers are live, the charts are reproducible from their links, and the datasets behind them are open to explore. Each post is explicit that it is a live demonstration of QuantumLayers, precisely because the series leans on the platform from the first chart to the last statistic.

Read the series so far: 3,023 Squirrels, One Park, One Census, 958 Volcanoes and 2.6 Billion Neighbours, What 45,716 Meteorite Landings Look Like, and What 10,877 Earthquakes in 30 Days Look Like. To load a public dataset of your own and get the charts and the findings the same way, start building at www.quantumlayers.com.