AQIF flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Central limit theorem

For independent identically distributed variables with finite variance, the sample mean converges to a normal distribution with mean μ and variance σ²/n as the sample size grows.

Type I and Type II error

Type I error is rejecting a true null hypothesis with probability α; Type II error is failing to reject a false null with probability β, so power equals 1 − β.

Adjusted R-squared

R² penalised for the number of regressors: 1 − [(1 − R²)(n − 1)/(n − k − 1)]. It falls when an added variable explains little incremental variation.

Heteroskedasticity- and autocorrelation-consistent errors

Newey–West or White standard errors correct inference when residual variance is non-constant or autocorrelated. OLS coefficients remain unbiased but conventional standard errors are invalid.

Stationarity

A time series whose mean, variance, and autocovariance do not depend on time. Regressing unrelated unit-root series on each other produces spurious correlation unless the data are differenced.

Modified duration

Macaulay duration divided by one plus yield per period. It approximates the percentage bond price change for a small parallel shift in yields.

Put-call parity

For European options on a non-dividend-paying stock, C − P = S − K·e^(−rT). A deviation from this identity allows a riskless arbitrage using a replicating portfolio.

Futures cost of carry

Fair futures price F = S·e^((r − q)T), where q is the dividend or convenience yield. The basis between futures and spot converges to zero at expiry.

ETF creation and redemption

Authorized participants exchange baskets of underlying securities for ETF shares and vice versa, arbitraging premiums or discounts to NAV so the market price tracks fair value.

Credit default swap

A bilateral contract in which the protection buyer pays a periodic spread and receives compensation if the reference entity experiences a defined credit event.

Bias-variance tradeoff

Expected squared prediction error equals bias squared plus variance plus irreducible noise. More complex models reduce bias but increase variance, raising out-of-sample error past an optimum.

L1 versus L2 regularization

Lasso (L1) shrinks coefficients and sets some exactly to zero, selecting features; ridge (L2) shrinks smoothly toward zero without eliminating variables from the model.

Purged k-fold cross-validation

A time-series validation scheme that drops training observations whose label windows overlap the test window and adds an embargo period, preventing leakage from autocorrelated outcomes.

Principal component analysis

Projects data onto orthogonal directions ordered by the eigenvalues of the covariance matrix. The first component explains the largest share of total variance.

Information coefficient

The correlation between forecast and realised returns, used to rank the predictive power of alpha signals. The Spearman rank version is robust to outliers.

Value at Risk

A loss threshold not expected to be exceeded with a stated probability over a given horizon. VaR is not subadditive, so portfolio VaR can exceed the sum of component VaRs.

Expected shortfall

The average loss conditional on exceeding the VaR threshold. It is a coherent risk measure because it is subadditive and captures tail severity beyond the quantile.

Risk parity

An allocation in which each asset contributes equally to total portfolio risk. It typically requires leveraging lower-volatility assets and differs from equal capital weights.

CAPM beta

Beta equals the covariance of an asset's return with the market return divided by the market variance, measuring sensitivity to market excess returns in a single-factor equilibrium model.

Kelly criterion

Optimal growth fraction f* = μ/σ² for continuous returns. Full Kelly maximises long-run compounding but produces severe drawdowns, so fractional Kelly is common in practice.

Square-root market impact

Expected execution price impact scales approximately with the square root of trade size relative to average daily volume, making trading costs concave in size and capacity finite.

Implementation shortfall

The gap between the paper portfolio return measured at the decision price and the actual realised return, capturing delay cost, market impact, and opportunity cost.

Alpha decay

The erosion of a signal's predictive power as information diffuses into prices. Faster decay forces higher turnover and tighter execution to capture the remaining edge.

Look-ahead bias

Using information that was not available at the simulated decision time. It inflates backtest results in ways that cannot be reproduced in live trading.

Survivorship bias

Building a historical universe from securities that still exist today. Excluding delisted or failed companies overstates historical returns and understates risk.

Deflated Sharpe ratio

A Sharpe ratio adjusted for the number of trials, skewness, kurtosis, and track-record length, correcting for selection effects when many configurations are tested.

Walk-forward analysis

Repeatedly optimising on an in-sample window and evaluating on the subsequent out-of-sample block, rolling forward in time to mimic live re-estimation of the model.

Material nonpublic information

Information that could reasonably move a security's price and is not publicly known. Trading on it, or tipping others, constitutes insider dealing under most regimes.

Mosaic theory

Assembling public information together with non-material nonpublic fragments to reach an investment conclusion is permissible, provided no piece is itself material or obtained in breach of duty.

Best execution

The duty to obtain the most favourable terms reasonably available for client orders, considering price, cost, speed, likelihood of execution and settlement, and order size.

Model risk governance

A framework of model inventory, independent validation, documented assumptions, assigned ownership, and periodic revalidation so model failures are detected and escalated promptly.

Spoofing

Entering orders with no intention to execute, then cancelling them, to create a false impression of supply or demand. It is prohibited market manipulation in most jurisdictions.