题目1707 · 统计
Why Hierarchical Testing Can Help
A researcher first tests whether a sector shows any effect, and only if that passes does she test stocks inside that sector. Why can this hierarchical design reduce the multiplicity burden?
打开 →找到 30 个结果
中文题目题目2716 · 机器学习
Why Stop-Loss Tuning Is Also Multiple Testing
Why does picking a stop-loss threshold after looking at the full historical equity curve count as backtest search rather than risk management hygiene?
打开 →题目3718 · 随机过程
Backtesting a Signal Versus Pricing Its Optional Overlay
A PM is backtesting a signal-driven strategy and also wants to value a stop-loss overlay that behaves like an option. Which measure should dominate each part of the workflow?
打开 →题目4937 · 数理金融
Backtesting Intuition 22
Why is ES harder to backtest directly than VaR in day-to-day risk control?
打开 →题目3719 · 随机过程
Stress Testing a Hedge Book Versus Marking It Today
Risk asks for a stress distribution of tomorrow's hedge-book PnL, while front office asks for today's mark of the same book. Should both use the same measure?
打开 →题目3486 · 数学
Why Adaptive Testing Beats Static Menus
Why can an adaptive questioning strategy distinguish more states than a non-adaptive test battery with the same primitive tests?
打开 →题目5281 · 金融与交易
Correlation Shock And Volatility Repricing 1
An equal-weight portfolio holds two assets with volatilities 0.04 and 0.09. The current correlation estimate is -0.2, but stress testing revises it to +0.3. With weights unchanged, what is the new portfolio volatility, and by how many volatility points does it rise versus the ori
打开 →题目1720 · 统计
Low Prior Probability and Positive Predictive Value
Suppose only 1% of tested trading ideas are genuinely predictive. A testing pipeline has 80% power and a 5% false-positive rate. Conditional on obtaining a positive result, what fraction of positives are truly real?
打开 →题目1685 · 统计
Power of a One-Sided z-Test
Consider testing $H_0:\mu=0$ versus $H_1:\mu>0$ with known $\sigma=10$ and sample size $n=25$ at significance level $\alpha=0.05$. What is the power when the true mean is $\mu=4$?
打开 →模块3.1.3 · 编程 · Python 基础
Python 打包与测试
python · packaging · pyproject · venv · uv · pip · dependencies · testing
打开 →模块3.2.3 · 编程 · Python 数据与量化分析
SciPy 与统计工具
python · scipy · stats · distributions · fit · var · descriptive-statistics · hypothesis-testing
打开 →模块2.6.5 · 数学与统计能力 · 机器学习理论
金融量化中的机器学习
machine-learning · financial-ml · cross-validation · purged-cv · cpcv · deflated-sharpe · multiple-testing · backtest-overfitting
打开 →课程Python 打包与测试 · Python 基础
pytest 基础
周一开盘前一刻钟，你在私募的研究服务器上 merge 了一段对 mean price 的「无害重构」——只是把 sum(...) / len(...) 拆成两步，方便在中间加日志。脚本照常跑完，回测照常出图。下午两点你才发现 PnL 报表上 XYZ001.SH 的当日均价对不上：你在重构时把 sum 与 len 的参数搞反了，函数对所有非空输入都返回 1 。...
打开 →课程Python 打包与测试 · Python 基础
覆盖率、基于属性的测试与 Mock
周五下班前你在私募的 CI 仪表盘上看到一片绿： xyzprice 的 86 个测试全过，行覆盖率显示 95%。周一开盘九点二十，研究系统在喂一段空盘后行情时崩在了 mean price([]) 上——你写过的测试里，从来没有一个把空列表喂进去。覆盖率告诉你「这一行跑过」，但不会告诉你「这一行只在 happy path 上跑过」。上一课的 pytest 让你...
打开 →题目1719 · 统计
0.049 vs 0.051 Decision Cliff
Two backtests differ only slightly: one reports p = 0.049 and the other p = 0.051. Why is it bad practice to call one ‘real’ and the other ‘not real’ purely because one is below 0.05?
打开 →题目2887 · 概率
A/B Gap Concentration
You run an A/B test with $n$ Bernoulli observations in treatment and $n$ in control, all independent. Let $\bar X$ and $\bar Y$ be the sample means. Use Hoeffding's inequality to bound \[ P\bigl((\bar X-\bar Y)-E[\bar X-\bar Y]\ge \varepsilon\bigr). \]
打开 →课程交易成本与市场冲击 · 回测与执行
Almgren-Chriss 成本模型
中信 CITIC 算法交易部一位资深执行交易员，正在与一家中型量化私募的投资经理通电话。私募需要在收盘前 30 分钟清掉 100 万股 600519 贵州茅台。到达价 RMB 1800.00；距离收盘 30 分钟。投资经理要求订单完成。交易员冷静地解释：「直接打盘口市价单，意味着接下来 30 秒 100% 参与率， 250 bp 冲击。30 分钟内做 TWA...
打开 →课程金融量化中的机器学习 · 机器学习理论
Alpha 信号与特征工程:标注、元标注与信号衰减
钩子:五十条弱 alpha 与一个总组合 你在一家中证500 中频量化私募(private fund)工作。研究团队在过去六个月里训练出了五十条独立的 ML alpha:有用 LightGBM 在 沪深300 / 中证500 因子风格暴露上做次日 alpha 的,有 1 D CNN 在分钟线上做日内动量(momentum)的,有 Transformer 在卖...
打开 →题目1697 · 统计
Any False Sector Winner After Within-Sector Mining
A desk has 12 sectors, each containing 5 genuinely null variants. In each sector it keeps only the smallest p-value, and it flags the sector if that winning p-value is below 1%. Assuming independence, what is the probability at least one sector is falsely flagged?
打开 →题目1689 · 统计
Approximate Fair-Coin Test with a Normal Approximation
A coin is flipped 20 times and lands heads 14 times. Use the normal approximation to test fairness at the 5% two-sided level.
打开 →题目1883 · 统计
Average Gross Leverage at Launch
A live manager panel shows 27 low-leverage funds and 18 high-leverage funds. Survival rates for those groups were 90% and 60%, respectively. Suppose low-leverage funds average 1.2x gross leverage and high-leverage funds average 2.4x gross leverage. What was the average gross lev
打开 →题目1711 · 统计
Backtest Slide Misreads p = 0.03
A PM deck says: ‘The event-study p-value is 0.03, so there is a 97% probability the signal is real.’ What is the statistical mistake?
打开 →题目1718 · 统计
Best-of-50 Reporting Without Adjustment
A researcher tests 50 candidate features and only reports the one with the smallest p-value, which happens to be 0.01. Why is it misleading to present 0.01 as if it came from a single pre-specified test?
打开 →题目1895 · 统计
Better Dataset for Failure-Rate Estimation
A vendor offers two hedge-fund datasets. Dataset A contains only funds that are currently reporting, but it includes long backfilled histories for those funds. Dataset B stores monthly reporting snapshots and preserves closed funds in the historical archive. Which dataset is be
打开 →题目1700 · 统计
BH q Range That Keeps Exactly Three Families
Five family-level winners have ordered p-values 0.004, 0.011, 0.018, 0.031, and 0.070. For what range of BH target levels q would Benjamini-Hochberg keep exactly the first three discoveries?
打开 →题目1698 · 统计
Bonferroni Cutoff After Collapsing to Effective Families
A research grid contains 60 model variants, but the desk argues they amount to only 15 effectively distinct families. If it wants family-wise error at most 10% using a Bonferroni family-level rule, what p-value cutoff should it apply to each effective family?
打开 →题目2702 · 机器学习
Chance the Best Null t-Statistic Exceeds 2.4 Across 50 Variants
Suppose 50 genuinely null standardized t-statistics are approximately independent N(0,1). What is the probability the largest of them exceeds 2.4?
打开 →题目1722 · 统计
Cherry-Picked Endpoint
A note tests 12 strategy diagnostics but highlights only the one with p = 0.02. What trap should the reviewer flag?
打开 →题目1684 · 统计
Chi-Square Goodness-of-Fit for a Die
A six-sided die is rolled 60 times, and the counts are $(14,8,11,9,10,8)$. Test fairness using a chi-square goodness-of-fit test.
打开 →题目1688 · 统计
Chi-Square Test for a Normal Variance
Assume normal data with $n=20$ and sample variance $s^2=9$. Test $H_0:\sigma^2=4$ using the classical chi-square variance test.
打开 →