Chapter 2 · 6 levels
Distributions and sampling
Read the shape of a column from its histogram, bucket values into bands, and learn how far a sample can sit from the population, with the standard error to say how far.
StartThe last chapter boiled a column down to a few numbers. This one looks at the whole shape behind them. A histogram shows where the values pile up, how long the tail is, and whether there is a spike nobody mentioned, like the thousands of one-page visits on Harbor & Pine’s site. A box plot puts several groups side by side, and you’ll learn to read one without the usual mistakes.
Then comes the question every analyst meets as soon as data gets big or expensive: what happens when you only look at part of it? A random sample of 200 orders gives a different average every time you draw it, and a sample picked the wrong way gives a wrong answer every time. You will measure both problems on the store data: random sampling error, which the standard error puts a number on, and bias, which no formula fixes.
The boss turns this into a planning answer: how many orders a survey or a test needs before its average is within $25 of the truth.
Levels
- How deep do visitors browse?
from Ashley Rivera, Web Content Lead
histogram, skew, distribution shape, bounce
- A sample of 200 orders
from Brian Mitchell, Finance Analyst
sample, population, sampling error, random sample
- The email-only estimate
from Lauren Hayes, Email Marketing Manager
sampling bias, selection bias, representative sample
- Read the box plot
from Megan Collins, Head of Sales
box plot, quartiles, iqr, outliers, comparing distributions
- Boss: how big should the sample be?
from Laura Bennett, Head of Finance
standard error, sampling distribution, simulation, sample size, margin of error