CommerceLens
Probability

Independence Check

Interactive

Two events A and B are independent when knowing one happened tells you nothing about the other. Formally, that means P(A ∩ B) should equal the product P(A) · P(B). We check that against the observed sample for several event pairs.

Three Required Comparisons

Discount Exposure × Purchase
Departs from independence in sample
A = Discount Exposure, B = Purchase
P(A ∩ B) — observed
15.682%
P(A) · P(B) — expected if independent
12.879%
P(A) = 46.770%P(B) = 27.537%Absolute difference = 0.028Relative difference = 21.8%
Mobile Device × Purchase
Appears independent in sample
A = Mobile User, B = Purchase
P(A ∩ B) — observed
17.980%
P(A) · P(B) — expected if independent
18.772%
P(A) = 68.172%P(B) = 27.537%Absolute difference = 0.0079Relative difference = 4.2%
Returning Customer × Purchase
Appears independent in sample
A = Returning Customer, B = Purchase
P(A ∩ B) — observed
25.277%
P(A) · P(B) — expected if independent
24.497%
P(A) = 88.962%P(B) = 27.537%Absolute difference = 0.0078Relative difference = 3.2%

Check Any Two Events

Interactive Independence Checker
Pick any two of the five named events and compute the same comparison live from the underlying session-level data.
Event A
Event B
Independence Test
P(P ∩ C) =? P(P) · P(C)
Purchase × Add to Cart
Departs from independence in sample
A = Purchase, B = Add to Cart
P(A ∩ B) — observed
13.978%
P(A) · P(B) — expected if independent
7.004%
P(A) = 27.537%P(B) = 25.435%Absolute difference = 0.0697Relative difference = 99.6%

Understanding Independence

Two behaviors are independent if seeing one gives you no extra information about the other. If mobile usage and purchasing were perfectly independent, the mobile purchase rate would equal the overall purchase rate exactly. In real data, small gaps between the observed and expected joint probability are normal even under independence, purely from sampling noise.

This is an empirical comparison under the observed dataset — a small absolute difference does not constitute a formal statistical proof of independence, and should not be presented as such. A proper test (e.g. a chi-squared test of independence) would need to account for sample size and sampling variability before drawing that conclusion.