Level II CFA Quantitative Methods and Economics Flashcards

Comprehensive vocabulary flashcards covering Level II CFA Quantitative Methods (Multiple Regression, Time-Series Analysis, Machine Learning, Big Data Projects) and Economics (Currency Exchange Rates, Economic Growth).

p-Value

The smallest level of significance for which the null hypothesis can be rejected.

Adjusted R2

A measure of regression goodness of fit modified for the number of independent variables, ensuring that adding a non-significant feature does not automatically increase the value.

Akaike's Information Criterion (AIC)

A goodness-of-fit metric used to evaluate competing models for the same dependent variable, preferred when the primary objective is generating better forecasts.

Schwarz's Bayesian Information Criterion (BIC)

A model selection metric that imposes a higher penalty for overfitting than AIC, preferred when the primary objective is identifying the model with the best goodness of fit.

Conditional Heteroskedasticity

A condition where the variance of regression residual terms is systematic and related to the levels of the independent variables, causing unreliable standard errors and invalid statistical tests.

Breusch-Pagan Test

A statistical test for conditional heteroskedasticity that regresses the squared residuals of an original model on its independent variables using a chi-square ($\chi^2$) test statistic.

Serial Correlation

A violation of regression assumptions where regression residual terms are correlated with one another across observations or time periods.

Multicollinearity

A condition in multiple regression where two or more independent variables are highly correlated with one another, inflating standard errors and lowering individual $t$-statistics.

Variance Inflation Factor (VIF)

A metric used to quantify the severity of multicollinearity for an independent variable $j$, calculated as $\frac{1}{1 - R_j^2}$, where values above 5 or 10 warrant concern.

Studentized Residual

A residual calculated by deleting an observation sequentially, estimating the model on $(n - 1)$ observations, and dividing the resulting forecast error by its standard deviation to identify outliers.

Logistic Regression (Logit) Model

A nonlinear regression model that transforms qualitative binary dependent variable probabilities into log odds to estimate discrete event likelihoods.

Covariance Stationarity

A time-series property defined by a constant and finite expected value, constant and finite variance, and constant and finite covariance between observations at any given lag.

Autoregressive (AR) Model

A time-series model structure where a dependent variable is regressed against one or more of its own lagged past values.

Mean Reversion

The tendency of a covariance-stationary time series to move back toward its long-run average level over time, calculated as $\frac{b_0}{1 - b_1}$ for an $AR(1)$ model.

Root Mean Squared Error (RMSE)

A criterion used to measure out-of-sample forecasting accuracy, equal to the square root of the average of squared forecast errors.

Unit Root

A condition in a time-series model where the coefficient on the lagged dependent variable equals 1 ($b_1 = 1$), causing the series to follow a nonstationary random walk process.

First Differencing

A transformation technique that converts a time series with a unit root into a covariance-stationary series by modeling changes in values ($y_t = x_t - x_{t-1}$).

Autoregressive Conditional Heteroskedasticity (ARCH)

A time-series condition where the variance of the error terms in one period depends on the magnitude of squared error terms from previous periods.

Cointegration

A long-term economic relationship between two nonstationary time series where a linear combination of the series produces covariance-stationary error terms.

Supervised Learning

Machine learning algorithms that utilize labeled training datasets containing defined target variables to train predictive classification or regression models.

Unsupervised Learning

Machine learning algorithms that process unlabeled datasets lacking target variables to discover hidden structures, groupings, or interrelationships among features.

Overfitting

A modeling problem in machine learning where excessive model complexity fits random sample noise, leading to high in-sample fit but poor out-of-sample generalization.

LASSO (Least Absolute Shrinkage and Selection Operator)

A penalized regression technique that minimizes sum of squared errors plus a penalty proportional to the sum of the absolute values of slope coefficients, effectively removing non-predictive features.

Support Vector Machine (SVM)

A linear classification algorithm that identifies an $n$-dimensional hyperplane separating data into binary classes while maximizing the margin to boundary observations.

K-Nearest Neighbor (KNN)

A nonparametric classification or regression algorithm that categorizes a new data point based on the majority class among its $k$ nearest historical observations.

Classification and Regression Trees (CART)

A decision tree model that recursively partitions feature spaces based on cut-off thresholds to predict binary/categorical outcomes (classification) or continuous variables (regression).

Principal Component Analysis (PCA)

An unsupervised dimension-reduction technique that transforms a large set of correlated features into a smaller set of uncorrelated factors called eigenvectors.

K-Means Clustering

An unsupervised algorithm that partitions data points into $k$ distinct, nonoverlapping clusters based on proximity to iteratively updated centroids.

Deep Learning Networks (DLN)

Neural networks composed of multiple hidden layers (typically 2 or more, and often over 20) used to model complex nonlinear relationships such as image or text recognition.

Data Wrangling

The process of cleansing, transforming, scaling, and structuring raw data to make it usable for statistical modeling and machine learning algorithms.

Normalization

A data scaling technique that rescales numerical feature values to lie strictly within a range between 0 and 1.

Standardization

A data scaling process that centers variable distributions at a mean of 0 and scales them in units of standard deviations.

One-Hot Encoding (OHE)

A feature engineering technique that converts a categorical variable into a series of binary (dummy) variables suitable for machine processing.

Bag-of-Words (BOW)

A text processing method that extracts and pools all individual words or tokens from a document without maintaining their original narrative sequence.

N-grams

Contiguous sequences of $N$ items or words extracted from text to preserve word order and contextual meaning in natural language processing.

Precision

A classification accuracy measure calculated as the ratio of true positive predictions to all predicted positives ($\frac{TP}{TP + FP}$).

Recall

A classification performance metric calculated as the ratio of true positive predictions to all actual positive cases ($\frac{TP}{TP + FN}$).

F1 Score

The harmonic mean of precision ($P$) and recall ($R$), calculated as $\frac{2 \times P \times R}{P + R}$.

Cross Rate

An exchange rate between two currencies that is calculated indirectly from their respective spot rates against a common third currency.

Covered Interest Rate Parity

An arbitrage-enforced parity relationship stating that forward currency premiums or discounts offset nominal interest rate differentials between two nations.

Uncovered Interest Rate Parity

An unarbitraged economic condition holding that the expected percentage change in spot exchange rates equals the nominal interest rate differential between two currencies.

International Fisher Effect

An economic relationship asserting that nominal interest rate differentials between two countries equal the difference in their expected inflation rates.

Relative Purchasing Power Parity

A theory holding that changes in foreign exchange rates over time will exactly offset inflation rate differentials between two economies.

FX Carry Trade

An investment strategy involving borrowing in a low-interest-rate funding currency and investing in a higher-yielding currency to capture interest differentials.

Mundell-Fleming Model

An economic model evaluating short-term impacts of monetary and fiscal policies on interest rates and currency exchange rates under varying capital mobility regimes.

Dornbusch Overshooting Model

A monetary exchange rate model assuming short-run price stickiness, causing exchange rates to temporarily overshoot long-run equilibrium PPP values after policy changes.

Potential GDP

The theoretical real output ceiling that an economy can achieve and sustain over the long run without triggering inflationary price pressures.

Capital Deepening

An increase in the capital-to-labor ratio ($\frac{K}{L}$) over time, depicted as movement along a fixed aggregate productivity curve.

Total Factor Productivity (TFP)

A scale factor in aggregate production functions reflecting technological progress that shifts the productivity curve upward, enhancing both labor and capital output.

Neoclassical Growth Theory

A growth theory positing that long-run equilibrium per capita growth depends strictly on technological progress and labor's share of output, with capital deepening producing only temporary gains.

Endogenous Growth Theory

A growth framework assuming that private investments in physical capital and human capital (R&D) generate societal technological spillovers and constant returns to capital, permanently raising growth rates.

Conditional Convergence Hypothesis

The proposition that per capita output levels of less-developed nations will converge toward developed nations only if they share identical savings rates, population growth, and production functions.