STEM University — Prove skills by building

Your Work

Data science output is a governed statistical claim: anomaly Z = (Y − μ̂)/σ̂, contribution Lift(c) = p_a(c)/p_b(c), and segment comparison Δ_M = M̄(A) − M̄(B). Each gate runs against the full event stream.

Date range

Anomaly detection

argmax |Z_k| day index (0..59)

Classification of that day

Percent difference PD = (Y − μ̂)/|μ̂|

Contribution analysis

Chrome lift = p_a(Chrome) / p_b(Chrome)

Top contribution driver browser

Segment comparison

Largest Venn region (Chrome vs Safari users)

Δ_M(Chrome, Safari) — mean revenue per visitor diff

p_Chrome(Electronics) − p_Safari(Electronics)

Statistical rules

Anomaly granularity G:

Holiday set (select all canonical holidays):

A real signal requires both magnitude (Gap ≥ τ) AND significance (|Z| ≥ z) — which?

DSA tuple components (select the 8 canonical):

Holiday adjustment purpose:

Allowed driver set formula:

Evaluation

Practice checks are unlimited and don’t count toward your grade. Final submit is one-time and generates your certificate.

Practice the workspace, ask Dr. Okoye anything, and submit when you're ready. The grade lives in the workspace, not the chat.