Glossary

Econometrics and Python glossary

Short definitions for econometrics vocabulary used across all Ceteris Lab modules, plus the programming, data, finance, and modeling language used in Fundamentals of Python for Financial Econometrics.

Module 0Module 1Module 2Module 3Module 4Module 5Module 6Module 7Module 8Financial Python

217 curated definitions across the flagship course and Modules 0-8.

Flagship course

Fundamentals of Python for Financial Econometrics

80 short definitions covering Python, data preparation, econometrics, financial time series, and machine learning.

Activation function: A transformation that introduces nonlinearity into a neural-network layer.
API: A defined interface through which software can request data or services from another system.
ARCH: A model in which current conditional variance depends on past squared shocks.
ARIMA: A time-series model combining autoregressive, differencing, and moving-average components.
Autocorrelation: Correlation between a time series and its own lagged values.
Bias: The systematic difference between an estimator's expected value and the true parameter.
Boolean: A logical value that is either True or False.
Broadcasting: NumPy's rules for applying operations to arrays with compatible shapes.
Classification: A prediction task in which the target belongs to one of a set of categories.
Cointegration: A stable long-run relationship among nonstationary time series.
Conditional: Code that chooses which instructions to run based on a True or False condition.
Confidence interval: A range of plausible parameter values constructed from an estimate and its uncertainty.
Convolution: A sliding weighted calculation used to detect local patterns in spatial or sequential data.
Cross-validation: Repeated fitting and evaluation across data folds to estimate out-of-sample performance.
CSV: A plain-text tabular file format in which values are separated by commas.
Data type: A category that determines how a value is stored and which operations it supports.
DataFrame: A rectangular Python data object with named columns and rows.
Design matrix: The numerical matrix of regressors supplied to a statistical model.
Dictionary: A Python collection that maps unique keys to values.
dtype: The NumPy or pandas label describing how values are stored, such as integer or floating point.
Dummy variable: A zero-one variable representing a category, condition, or group.
Estimator: A rule or formula that uses sample data to estimate an unknown population parameter.
Exception: A Python object that signals an error or unusual condition during execution.
Feature: An input variable used by a statistical or machine-learning model.
Function argument: A value supplied to a function when the function is called.
GARCH: A volatility model in which conditional variance depends on past shocks and past variance.
GroupBy: A pandas operation that splits data into groups, applies calculations, and combines the results.
Histogram: A chart that groups numerical observations into intervals to show their distribution.
Interaction term: A variable created by multiplying variables to let one effect depend on another variable.
Interpreter: The program that reads and executes Python instructions.
JSON: A text format for exchanging structured data using objects, arrays, names, and values.
Kernel: The active computing process that runs code in a notebook.
Lag: A previous-period value of a time-series variable.
Lasso: A regression method that uses an absolute-value penalty and can set some coefficients to zero.
List: An ordered, changeable Python collection that can contain repeated values.
Log return: The natural logarithm of the ratio of an asset's ending price to its starting price.
Logistic regression: A model that links predictors to the probability of a binary outcome.
Loop: Code that repeats an instruction for items or while a condition remains true.
Merge: An operation that joins tables by matching one or more key columns.
Missing value: A value that is absent, unknown, or not recorded in a dataset.
Monte Carlo simulation: Repeated random simulation used to approximate a distribution, probability, or expected value.
Neural network: A layered prediction model that combines weighted inputs with nonlinear transformations.
None: Python's special value for the absence of a value.
Notebook: An interactive document that combines executable code, output, equations, and written explanation.
NumPy array: A fast, fixed-type structure for numerical vectors, matrices, and higher-dimensional data.
OLS: Ordinary Least Squares, which estimates coefficients by minimizing the sum of squared residuals.
Outlier: An observation that lies unusually far from the main pattern of the data.
Overfitting: Learning sample noise so closely that performance deteriorates on new data.
Package: A reusable collection of Python modules built for a particular set of tasks.
Python: An open-source programming language widely used for data analysis, econometrics, finance, and machine learning.
Python function: A reusable block of code that accepts inputs and may return an output.
Python module: A Python file containing reusable functions, classes, or variables.
Python variable: A named reference to a value or object in a Python program.
Random seed: A fixed starting value that makes a pseudo-random calculation repeatable.
Regularization: A penalty that discourages overly complex models and can improve out-of-sample performance.
Relative path: A file location written in relation to the current project or working directory.
Reproducibility: The ability to repeat an analysis and obtain the same results from the same inputs.
Return: The gain or loss on an asset relative to its initial value over a stated period.
Return value: The result that a function sends back to the code that called it.
Ridge regression: A regression method that shrinks coefficients using a squared-magnitude penalty.
Scatter plot: A graph of paired numerical values used to inspect association, clusters, and unusual observations.
Script: A saved Python file containing instructions that can be run again.
Series: A one-dimensional pandas object with values and an index.
Set: An unordered Python collection of unique values.
Shape: The number of elements along each dimension of an array or table.
Simulation: A computer-generated experiment used to study uncertainty or model behavior.
Slice: A selection of part of an ordered object using start, stop, and optional step positions.
Standard error: An estimate of how much a statistic would vary across repeated samples.
Stationarity: A time-series property in which key statistical features remain stable over time.
statsmodels: A Python package for statistical models, estimation, hypothesis tests, and diagnostics.
Target: The outcome a predictive model is trained to estimate or classify.
Tidy data: Data organized so each variable is a column, each observation is a row, and each value is a cell.
Time series: Observations ordered through time, often at regular intervals.
Train-test split: A division of data into one sample for fitting and another for evaluating predictions.
Tuple: An ordered Python collection whose items cannot be changed after creation.
Unit root: A source of persistent nonstationarity in which shocks do not quickly fade.
Unit test: A small automated check that verifies one part of a program behaves as expected.
Vectorization: Expressing a calculation as array operations instead of explicit Python loops.
Virtual environment: An isolated Python setup that keeps a project's packages and versions separate.
Volatility: The degree of variation in returns, commonly summarized by a standard deviation.

Modules 0-8

Econometrics module terms

Absolute value: The distance of a number from zero, written without regard to sign.
Adjusted R-squared: A goodness-of-fit measure that adjusts R-squared for the number of regressors in the model.
Alternative hypothesis: The claim considered when the null hypothesis is not supported by the evidence.
Argument: A reasoned claim supported by assumptions, logic, or evidence.
Asymptotic bias: The limiting gap between an estimator and the true population value.
Asymptotic efficiency: A large-sample comparison of estimator precision.
Asymptotic normality: The property that a properly scaled estimator is approximately normal in large samples.
Asymptotic standard error: A standard error based on large-sample approximation.
Average: A typical value of a variable, often represented by the sample mean.
Bad control: A control variable that can distort the research question or absorb part of the effect of interest.
Base group: The omitted category used as the reference point for interpreting categorical indicators.
Binary dependent variable: An outcome that takes only two values, often coded zero and one.
Binary variable: A variable coded as 1 when a condition is true and 0 otherwise.
BLUE: Best Linear Unbiased Estimator under the Gauss-Markov assumptions.
Breusch-Pagan test: A diagnostic test that relates squared residuals to variables that may explain error variance.
Causal claim: A statement that changing one factor would change an outcome.
Causality: A cause-and-effect relationship where changing one factor changes another.
Centering: Subtracting a reference value, often the mean, to make interactions or intercepts easier to read.
Central Limit Theorem: A result explaining why many averages and estimators are approximately normal in large samples.
Ceteris paribus: All else equal; comparing one factor while holding relevant factors fixed.
Chow test: A joint test for whether regression coefficients differ across groups.
Coefficient: An estimated number multiplying a variable in a regression equation.
Complement: The event that a stated event does not happen.
Confounder: A third factor related to both the explanatory variable and the outcome.
Consistency: The property that an estimator gets closer to the true parameter as sample size grows.
Control group: The group used as the comparison benchmark in a treatment or policy setting.
Control variable: A variable added to account for another relevant factor in the comparison.
Conventional standard error: A standard error computed under the usual homoskedasticity formula.
Correlation: A standardized measure of linear association between two variables.
Covariance: A measure of whether two variables tend to move above or below their means together.
Critical value: A cutoff from a reference distribution used in a hypothesis test.
Cross-sectional data: Data on many units observed at one point in time.
Degrees of freedom: The amount of independent information left after estimating model parameters.
Dependent variable: The outcome variable the model is trying to explain or predict.
Descriptive relationship: An observed sample pattern that does not by itself establish cause and effect.
Distribution: A description of possible values of a variable and how frequently or likely they occur.
Dummy variable trap: Perfect multicollinearity caused by including every category indicator along with an intercept.
Econometrics: The use of data and statistical methods to study economic relationships.
Elasticity: The percent change in an outcome associated with a one percent change in an explanatory variable.
Endogeneity: A problem where an explanatory variable is related to the error term.
Equation: A statement that two mathematical expressions are equal.
Error term: The part of the outcome not explained by the variables included in the model.
Exact percent change: A percent interpretation for log-dependent models that uses exponentiation rather than a rough approximation.
Exogeneity: A condition that the error has no systematic relationship with the explanatory variables.
Expected value: The long-run average value of a random variable.
Explanatory variable: A variable used to explain or predict the dependent variable.
F statistic: A test statistic used for joint hypotheses or overall regression significance.
Feasible GLS: A weighted estimator that first estimates the variance pattern and then uses predicted weights.
Finite-sample property: A property describing an estimator's behavior for a specific sample size.
Fitted line: The line chosen by regression to summarize the sample relationship.
Fitted value: The predicted value from an estimated regression equation.
Function: A rule that maps one or more inputs to an output.
Functional form: The mathematical shape used to connect the dependent variable to regressors.
Gauss-Markov assumptions: Conditions under which OLS is the best linear unbiased estimator.
Group mean difference: The difference in average outcomes between two groups.
HC standard error: A family of heteroskedasticity-consistent standard errors such as HC0, HC1, HC2, and HC3.
Heteroskedasticity: A condition where the error variance changes across observations, groups, or regressor values.
Heteroskedasticity-robust inference: Tests and intervals that use standard errors valid under changing error variance in large samples.
Holding other factors fixed: Interpreting one coefficient while keeping the other regressors unchanged.
Homoskedasticity: A condition where the error variance is constant across observations or regressor values.
Index: A counter or label that identifies an observation, variable, group, or time period.
Intercept: The predicted value of the dependent variable when explanatory variables equal zero.
Intercept shift: A change in the intercept associated with a binary or categorical variable.
Joint effect: The combined predicted change from changing more than one explanatory variable.
Joint significance: Evidence that a group of coefficients matters collectively.
Known-form heteroskedasticity: A case where the error variance pattern is known or defensibly specified.
Lagrange Multiplier test: A test that uses a restricted model to check whether excluded information matters.
Large-sample inference: Inference based on approximations that improve as sample size grows.
Law of large numbers: A result showing that sample averages converge to population averages under suitable conditions.
Level-log model: A model where the dependent variable is in levels and an explanatory variable is logged.
Linear probability model: An OLS model used when the dependent variable is binary.
Linear restriction: A hypothesis that places a linear condition on one or more coefficients.
Log-level model: A model where the dependent variable is logged and an explanatory variable is in levels.
Log-log model: A model where both the dependent variable and an explanatory variable are logged.
Mean: The arithmetic average of a set of values.
Median: The middle value after observations are sorted.
Multicollinearity: A strong linear relationship among explanatory variables that can reduce estimation precision.
Multiple regression: A regression model with one dependent variable and two or more explanatory variables.
Natural logarithm: A transformation often used to interpret changes in approximate percentage terms.
Nonnested models: Models where one specification is not a simple restricted version of the other.
Normal distribution: A symmetric bell-shaped distribution described by a mean and standard deviation.
Null hypothesis: The benchmark claim tested against the data, often that a coefficient equals zero.
Observation: One row, case, person, firm, region, or time period in a dataset.
Omitted base category: The category left out so included dummy coefficients are interpreted relative to it.
Omitted variable bias: Bias from leaving out a relevant variable that is correlated with an included regressor.
Over-control: Adding controls that block or distort the relationship the analysis is trying to measure.
p-value: A measure of how unusual the sample evidence would be if the null hypothesis were true.
Panel data: Data that follow the same units across multiple time periods.
Panel unit: The person, firm, region, or country followed over time in panel data.
Parameter: An unknown population value that a model tries to estimate.
Partial effect: The relationship between one explanatory variable and the outcome, holding other included variables fixed.
Partialling out: A way to isolate the part of a variable not explained by other controls.
Policy indicator: A binary variable that marks exposure to a policy or program.
Pooled cross section: Separate cross-sectional samples combined across multiple time periods.
Population: The full group of units or outcomes the research question is about.
Practical significance: Whether an estimated effect is economically meaningful in size, not only statistically detectable.
Prediction interval: A range that accounts for uncertainty when predicting an individual outcome.
PRF: Population regression function, the average relationship in the full population.
Probability: A number from 0 to 1 that describes how likely an event is.
Probability limit: The value an estimator converges toward as sample size grows.
Python package: A reusable collection of Python tools for data work, statistics, or visualization.
Quadratic term: A squared variable used to allow a curved relationship.
R-squared: The share of total variation in the dependent variable explained by the regression.
Random variable: A variable whose value is uncertain before it is observed.
Regression: A method for estimating relationships between an outcome and one or more explanatory variables.
Regression equation: A compact model that links an outcome to explanatory variables, coefficients, and an error term.
Repeated survey: A survey collected in multiple periods, often with new respondents each time.
Residual: The observed value minus the fitted value.
Residual sum of squares: The sum of squared residuals left unexplained by the model.
Robust F test: A joint test that uses a heteroskedasticity-robust covariance matrix.
Robust standard error: A standard error designed to remain valid under heteroskedasticity in large samples.
Robust t test: A t test computed with a heteroskedasticity-robust standard error.
Sample: The observations actually used in the analysis.
Sample size: The number of observations in the sample, often written as n.
Sampling distribution: The distribution an estimator would have across repeated samples.
Self-selection: A problem where people or units choose into treatment or groups in ways related to outcomes.
Simple regression: A regression model with one dependent variable and one explanatory variable.
Skewness: A measure of asymmetry in a distribution.
Slope: The predicted change in the dependent variable for a one-unit increase in an explanatory variable.
Slope shift: A change in a slope created by an interaction term.
Special White test: A version of the White test that uses fitted values and squared fitted values in the auxiliary regression.
SRF: Sample regression function, the fitted relationship estimated from sample data.
Standard deviation: The square root of variance, measured in the original units of the variable.
Standardized coefficient: A coefficient measured in standard-deviation units for comparison across variables.
Statistical significance: Evidence strong enough to reject a null hypothesis at a chosen significance level.
Summation: A compact instruction to add values across observations.
t statistic: A coefficient estimate divided by its standard error.
Time-series data: Data on one unit or variable observed across multiple time periods.
Treatment group: The group exposed to a policy, intervention, or condition of interest.
Turning point: The value where a quadratic relationship changes direction.
Two-sided test: A test that checks for evidence against the null in either direction.
Variable: A measured item that can change across people, places, firms, or time.
Variance: The average squared distance of values from their mean.
Variance function: A model or description of how error variance changes with variables.
Variance inflation factor: A diagnostic that summarizes how much multicollinearity inflates a coefficient's variance.
Weighted least squares: A least-squares method that gives lower weight to observations with higher error variance.
White test: A broad heteroskedasticity diagnostic using squares and interactions of regressors.