Welcome

What are you trying to do?

Something I need to watch regularly in government, parastatals and regulators.

Before you start. Monitors currently run from the command line rather than from a page. The engine works and is tested; the interface for setting them up is still being built, so this route is the least finished of the four.

Your path

Create the study

One study holds the whole series.

Chart it before you judge it

A control chart separates ordinary variation from a real change. Most month-to-month movement is noise, and acting on it makes a process less stable rather than more.

Set up a monitor

A saved analysis with a cadence. It re-runs and reports what moved, without claiming the movement is significant - two consecutive estimates are not independent samples.

Read the series, not the last point

Twelve points tell you what one never can.

What people in your field ask

In government, parastatals and regulatorsthe full guide

State the decision - assemble the administrative data - test whether the pattern is real - publish with the method attached.

Is the indicator actually moving, or is it noise?control chart
Are we improving over time?arima
Do states, LGAs or offices differ?one way anova
Where is the problem concentrated?morans i
Several units over several years - what explains the difference?panel regression

Watch for: Publishing a figure without the method attached. Every result here carries the digest of the data behind it, which is what turns a contested statistic into a checkable one.

Create your first study The full guide for government, parastatals and regulators Start again