1/107
Looks like no tags are added yet.
Expected value E(X)
The long-run average of a random variable.
Variance
How spread out values are around the mean; the average squared distance from the mean.
Standard deviation (σ)
The square root of the variance.
Covariance
Measures whether two variables move together. Positive = move in the same direction; negative = opposite directions. Depends on units.
Correlation
A standardized (unitless) version of covariance, always between −1 and 1.
Independence
Knowing one variable tells you nothing about the other. Implies zero covariance (but zero covariance doesn't imply independence).
Normal distribution
A symmetric, bell-shaped distribution described by its mean μ and variance σ².
z score
How many standard deviations a value is from the mean.
Unconditional expectation
The average of a variable across the whole population, E(Y).
Conditional expectation
The average of a variable within a subgroup, E(Y | X = x).
Random variable
A variable whose value is uncertain before it is observed.
Experimental data
Data where x is randomly assigned by the researcher.
Non-experimental data
Data where x is observed, not assigned; people choose their own x.
Micro data
Data on individuals, households or firms.
Macro data
Data on whole economies (GDP, inflation, unemployment).
Stock variable
A variable measured at a point in time (e.g., savings on Dec 31).
Flow variable
A variable measured per period of time (e.g., annual income).
Quantitative data
Data in numbers (wage, age).
Qualitative data
Data in categories (gender, region).
Cross-sectional data
Many units observed at one point in time.
Time series data
One unit observed over many time periods.
Repeated cross-sectional data
Different units sampled in each time period.
Panel (longitudinal) data
The same units followed over time.
Balanced panel
A panel where every unit is observed in every period.
Unbalanced panel
A panel where some units are missing in some periods.
Econometrics
Statistical tools used to understand causal relationships between variables using data.
Causality
x actually produces a change in y, not just a correlation between them.
Econometric model
An equation relating y to x made of parameters, variables and a random error.
Parameters
The unknown true values in a model (β1, β2).
Dependent variable (y)
The outcome being explained. Also called explained, response, predicted or regressand variable.
Independent variable (x)
The variable used to explain y. Also called explanatory, control, predictor or regressor variable.
Systematic part
The part of y explained by x: β1 + β2x.
Unsystematic part / error term (e)
The part of y not explained by x.
Linear regression model
A model that is linear in its parameters (β's), even if nonlinear in x or y.
Simple regression
A regression with one x variable.
Multiple regression
A regression with more than one x variable.
Endogeneity
When x is correlated with an omitted factor that also affects y.
Endogenous variable
The x variable in the regression that is correlated with an omitted factor.
Omitted / other / confounding variable
A factor left out of the regression that is correlated with both x and y.
Observable variable
A variable available in your dataset (e.g., age).
Unobservable variable
A variable you can't measure in your data (e.g., ability, motivation).
Omitted variable bias (OVB)
The bias in an estimate caused by leaving out a variable correlated with both x and y.
Positive OVB
The estimated effect is overstated (too high).
Negative OVB
The estimated effect is understated (too low).
Universally true (causality challenge)
The concern that a relationship found in one sample may only hold for that time or group of people.
Population
Information on everyone. Never available.
Sample
Information on a subset of the population. Always what we have.
Population regression model
y = β1 + β2x + e; the true relationship.
Population regression line
E(y|x) = β1 + β2x; the average y at each x.
Population error (e)
The gap between y and E(y|x). Never observed.
Sample regression line
ŷ = b1 + b2x; the line estimated from data.
Sample regression model
y = b1 + b2x + ê.
Estimates (b1, b2)
Values of the parameters calculated from a sample.
Intercept
The predicted y when x = 0 (β1 in the population, b1 in the sample).
Slope coefficient
The change in y for a one-unit change in x (β2 or b2).
Predicted / fitted value (ŷ)
The value of y on the regression line for a given x.
Residual (ê)
Actual minus predicted value: y − ŷ.
Measurement error
When a variable is recorded inaccurately.
Functional form error
Using the wrong shape for the relationship (e.g., a line for a curve).
Homoskedasticity
The spread of the errors is the same at every value of x (assumption #2).
Heteroskedasticity
The spread of the errors changes with x; a fan shape in a scatterplot.
Autocorrelation
Errors linked across time periods (common in time series).
Clustering
Errors linked within groups such as schools, families or cities.
Normality assumption
The errors are normally distributed (assumption #4).
Skewness
Asymmetry of a distribution. 0 = symmetric.
Right (positive) skew
A long tail to the right (e.g., income).
Left (negative) skew
A long tail to the left.
Kurtosis
Thickness of a distribution's tails. Normal = 3; above 3 = thicker tails; below 3 = thinner.
Outlier
An observation far away from the rest of the data.
Topcoding
Recording all values above a cap at the cap, creating a spike at the maximum.
Robust standard errors
Standard errors corrected for heteroskedasticity/clustering; the coefficients don't change.
hettest
A Stata test for heteroskedasticity; its null hypothesis is homoskedasticity.
Ordinary least squares (OLS)
The method that picks the line minimizing the sum of squared residuals.
SSE (sum of squared errors)
Unexplained variation in y: Σê². Stata's Residual SS.
SST (total sum of squares)
Total variation in y around its mean. Stata's Total SS.
SSR (regression sum of squares)
Variation in y explained by the regression. Stata's Model SS.
R-squared (R²)
The share of the variation in y explained by x.
σ̂² (estimated error variance)
SSE/(N − 2). Stata's Residual MS.
Root MSE (σ̂)
The standard error of the regression; the square root of σ̂².
Sample variance of y (s²ᵧ)
SST/(N − 1). Stata's Total MS.
Goodness of fit
How well the regression line matches the data (judged by R², SSE, σ̂).
Dummy (indicator) variable
A variable equal to 0 or 1 representing a category. Named after the category coded 1.
Base category
The category coded 0 in a dummy variable.
Dummy variable trap
Including a dummy for every category along with an intercept; you must leave one out.
Linear probability model (LPM)
An OLS regression where y is a dummy; coefficients are changes in the probability that y = 1.
Functional form
The shape of the relationship between x and y in a model.
Linear model
Both x and y in levels; a straight-line relationship.
Quadratic model
y regressed on x²; captures U-shaped relationships.
Cubic model
y regressed on x³; captures relationships switching between concave and convex.
Log-lin model
log y regressed on x.
Log-log model
log y regressed on log x.
Lin-log model
y regressed on log x.
Coefficient
The estimated b2 for x in any model.
Slope
dy/dx: the change in y (units) for a one-unit change in x.
Elasticity (ε)
The % change in y for a 1% change in x. Unitless.
Semi-elasticity (η)
The % change in y for a one-unit change in x.
Evaluated at the mean
Calculating slope or elasticity using the mean values of x and y.
des
Stata command that describes the variables in a dataset.
gen
Stata command that creates a new variable.
replace
Stata command that changes the values of an existing variable.