CSV statistics: mean, standard deviation, min and max

The wind speed counts 8,736 values and 24 missing ones: the 24 hours of 12 March. Its mean and standard deviation leave them out, rather than counting them as zero. The rainfall’s median and quartiles are 0: at least three hours in four are dry. Zoom the chart into January and every card shows January’s numbers instead.

The same in Python

With the sample file downloaded, these lines give pandas’ answer (skiprows=[1] skips the units row; pandas’ std divides by n − 1, as the panel does, and its quantile interpolates as the panel’s Q1, median and Q3 do):

import pandas as pd

d = pd.read_csv("megarows-sample-leeds-hourly-weather.csv", skiprows=[1])
x = d.drop(columns="time")
table = pd.DataFrame({"count": x.count(), "mean": x.mean(), "std": x.std(),
                      "min": x.min(), "max": x.max(), "rms": (x**2).mean() ** 0.5})
print(table.round(3).to_string())
print(x.quantile([0.25, 0.5, 0.75]).T.round(3).to_string())

They print, rounded to three decimals:

                   count    mean    std    min     max     rms
Leeds_Temperature   8760   9.566  5.208  -3.33   21.88  10.891
Leeds_Humidity      8760  80.058  8.303  53.00  100.00  80.487
Leeds_Wind_Speed    8736  10.755  4.783   0.00   28.60  11.771
Leeds_Rainfall      8760   0.068  0.293   0.00    4.30   0.301
                     0.25   0.50    0.75
Leeds_Temperature   5.708   9.98  13.402
Leeds_Humidity     74.000  80.00  86.000
Leeds_Wind_Speed    7.400  10.60  13.900
Leeds_Rainfall      0.000   0.00   0.000

We ran them on 9 October 2026 with numpy 2.1.3, pandas 2.2.3, Python 3.12.15. A test works out the panel’s numbers on every change to the site and checks that they are the ones on this page and that they match NumPy’s to about 12 significant digits. Download the sample file (CSV, 299 KB).

What each statistic means

Count

The values that are there; empty cells are left out of every other statistic.

Missing

Empty cells, and markers such as NA or n/a, with their share of the rows.

Min and Max

The smallest and the largest value.

Mean

The sum of the values divided by the count.

Median

The middle value once the values are sorted; with an even count, halfway between the middle two. Half the values are below it, half above.

Std dev

The sample standard deviation: the square root of the sum of squared differences from the mean, divided by n − 1, as MATLAB’s std and Excel’s STDEV. 0 for a single value.

RMS

The root mean square: the square root of the mean of the squared values. For a signal that swings around zero, its typical size.

Peak to peak

Max minus min.

Q1 and Q3

The quartiles: a quarter of the values lie below Q1 and a quarter above Q3. Between two values they are interpolated linearly, as NumPy, pandas and Excel’s QUARTILE.INC do; MATLAB’s quantile uses another rule, so for a few values its Q1 and Q3 can differ slightly (the median is the same).

IQR

The interquartile range, Q3 minus Q1: the spread of the middle half of the values, which a few extreme values do not move.

For the rows in view, the whole file or between cursors

The panel has three choices. Visible range, the default, covers the rows the chart shows: drag across the chart to zoom into a part (or use the scroll wheel, or a pinch on a touch screen), and the numbers follow once the chart rests. Whole file covers every row, whatever the zoom. Between cursors appears when the difference cursors are on (the cursor button over the chart, or the D key): it covers the rows from cursor C1 to C2, which you can drag to an exact time.

Whatever the choice, every value in the range is used at full precision, millions of rows too: the chart may draw an overview of a long file, but the statistics never come from it.

Text and category columns

Columns of text, such as a station name or a status, have no mean. Choose Show statistics in the column’s menu in the table: the panel shows its count, missing values, the number of distinct values and the most common ones, over the whole file.

Checked against NumPy

Private by design: nothing is uploaded

Measurement and business data are often confidential. This tool reads your file and works out the statistics in your browser, so the file and its name never leave your device. Files with millions of rows work: the practical limit is your device’s memory.