Mathematics Interview Questions: Theorem-Proof Drilling and Analysis Foundations

What this page helps you do first

• The signature round: narrating a proof live
• Epsilon-delta precision in definitions is the floor
• Academic and applied-statistics tracks prepare differently

Drilling chains for the three core courses

| Course | Entry question | Follow-up path | | :--- | :--- | :--- | | **Analysis** | State and prove the nested-interval theorem | Equivalence with the supremum principle → proving Bolzano-Weierstrass with it → the completeness family’s mutual derivations | | **Algebra** | Necessary and sufficient conditions for diagonalizability? | Geometric vs algebraic multiplicity → motivation for Jordan form → similarity vs congruence | | **Probability** | LLN vs CLT? | Convergence in probability vs distribution → the i.i.d. role → one application scene |

The signature of mathematics drilling is **chained equivalence**: panels happily follow “how else can this be proved / how does it relate to theorem XX”. Prepare relation webs — who proves whom, what is equivalent, where each applies — rather than isolated theorems.

Handling the live-proof round

Three layers of response: **frame first** (“by contradiction: assume existence, construct a nested sequence, derive a contradiction”) — panels will prompt along your frame; **step back when stuck** (“the constant in this estimate escapes me, but the goal is bounding XX”) — correct direction earns partial credit; **mark honestly** (“I recall this lemma imprecisely; roughly it says …”). Panels hold no obsession with remembered inequalities and enormous regard for structural proof sense.

The must-drill list: nested intervals, supremum principle, the mean value theorem, Newton–Leibniz, rank inequalities, Bolzano–Weierstrass — each to three-minute-frame plus key-steps level.

Definitional precision is the floor

“Roughly correct” definitions count as wrong. High-frequency recitations: the epsilon-delta definition, uniform vs pointwise convergence, linear dependence, measure, the axioms of probability. The discipline: **quantifier order first** (“for every epsilon>0 there exists delta>0 such that …”) — reversed quantifiers reverse the definition. Drill: write ten core definitions from memory three times, checking quantifier nesting each pass.

Academic vs applied-statistics divergence

**Academic mathematics**: weight on analysis and algebra proofs plus a first-pass acquaintance with the target direction (equations/geometry/topology — one paragraph each of why interested). **Applied statistics**: weight on statistical intuition (the logic of hypothesis testing, a plain-language p-value, regression diagnostics) plus tooling evidence (a small R/Python analysis). Applied candidates should not sink time into measure-theoretic rigor — panels want statistical thinking plus tools.

Sprint checklist

• **Relation webs**: one per core course;
• **Six oral proofs**: three-minute frames, recorded and reviewed;
• **Ten definitions written**: quantifier order checked verbatim;
• **Direction statement**: one interest paragraph (what you read, what you want to ask);
• **Applied add-on**: one R/Python case narrated end-to-end — cleaning, modeling, diagnostics.

Frequently asked questions

Will freezing mid-proof end the assessment?

No — panels widely treat post-freeze behavior as more informative than the proof itself. The move: name the stuck step (“the constant in the estimate escapes me”), state the step’s goal (“this term must be bounded by epsilon/3”), and ask to continue. Structural sense plus metacognition routinely outperforms fluent recitation.

My undergraduate record is mediocre — how to survive heavy theorem drilling?

Contract the perimeter: polish the most central clusters (analysis: completeness plus the mean value theorem; algebra: diagonalization plus rank; probability: the two limit theorems) to chain-proof level, holding concept-level answers elsewhere. Deep probing locates your ceiling; surrender unreachable ceilings honestly — but never lose base points on definitions, which signal the floor.

Cross-applying from mathematics to CS/finance — how to position the background?

Mathematics is hard currency in both directions; the task is a positive narrative for “why leave”: “not leaving mathematics but moving it closer to problems” — statistics for data science, stochastic processes and optimization for finance. Expect depth probing on probability and statistics (near-certain), presenting your two strongest courses as the evidence chain. The formula: mathematical foundation plus concrete interest evidence in the target field (courses, projects, books).

Where to go after this question bank

Question banks rehearse the follow-up chains; your own materials decide whether the answers hold. Use the thesis workflow to strengthen the draft behind your answers.