Statistics & drivers
What goes with what, tested properly.
Describe a distribution, test whether two groups differ, find what moves together, smooth a trend or split a change into its contributions — with scipy and statsmodels underneath, and corrections when many tests run at once.
- Difference
- +0.31
- Effect size
- 0.41
- p, corrected
- 0.001
People who completed a course this year are rated higher — a moderate difference, unlikely to be chance after correcting for the other tests run.
Statistics & drivers, part by part.
- 01
Describe and compare
The shape of a measure, and whether two or more groups really differ.
- 02
Associations
Correlations between measures, and chi-square tests between categories.
- 03
Trends and contributions
Moving averages, and how much each part contributed to a change.
- 04
Corrected for many tests
When many tests run at once, the results are corrected (Holm, Benjamini–Hochberg), so chance findings are not called real.
Pick, Test, Correct, Explain.
- Step 1
Pick
The measures and the groups.
- Step 2
Test
With the method that suits the data.
- Step 3
Correct
For running many tests at once.
- Step 4
Explain
The result, in words.
Six tests the analyst can run
- Describe
- How is tenure spread across the company?
- Correlation
- Does time since the last raise go with lower ratings?
- Chi-square
- Is exit reason related to department?
- Compare means
- Are people who completed a course rated higher?
- Moving average
- What is the underlying trend in monthly exits?
- Contribution
- Which departments drove the change in headcount?
Questions it answers, in your words.
- “Are people who completed a course this year rated higher?”
- “Is exit reason related to department?”
- “Which departments contributed most to the rise in attrition?”
On the same model and graph
- Deterministic factsEvery figure computed by the KPI engine, a query on the graph or a statistics library — never made up by a model.
- Root cause analysisWhere a KPI moved and why: the change split, causes searched in your own graph, each hypothesis tested.
- Forecasts & predictionsForecasts with ranges you can trust, and predictive models you can read.
Your next people decision, with the why attached.
Request a demo. We'll email you a two-minute walkthrough today, then set up a time that suits you.