1/49
Comprehensive practice vocabulary flashcards covering simple and multiple linear regression models, least squares estimation, diagnostic residual evaluation, categorical and nonlinear modeling, data wrangling methodologies, missing data taxonomy, and hypothesis testing principles.
Regression Analysis
A statistical procedure that uses data to develop an equation showing how variables are related.
Dependent Variable
The response variable being predicted in a regression analysis, denoted by y.
Independent Variable
A predictor variable or feature used to predict the value of the dependent variable, denoted by x (or x1,x2,…,xq).
Simple Linear Regression Model
An equation describing how the dependent variable y relates to a single independent variable x and an error term ϵ, formulated as y=β0+β1x+ϵ.
Multiple Linear Regression Model
A regression model relating the dependent variable y to two or more independent variables through the equation y=β0+β1x1+β2x2+⋯+βqxq+ϵ.
Error Term (ϵ)
The model component that accounts for the variability in the dependent variable y that cannot be explained by the linear effect of the independent variable(s).
Least Squares Method
A procedure for using sample data to determine the estimated regression equation by minimizing the sum of squared residuals, ∑(yi−y^i)2.
Residual (ei)
The error made in estimating the value of the dependent variable for the ith observation, calculated as the observed value minus the predicted value: ei=yi−y^i.
Experimental Region
The range of values of the independent variables in the sample data used to estimate the regression model.
Extrapolation
The prediction of the value of the dependent variable outside the experimental region, which is risky and unreliable unless supported by empirical evidence.
Sum of Squares Due to Error (SSE)
A measure of the error resulting from using the predicted values y^i to estimate the actual values yi, computed as SSE=∑(yi−y^i)2.
Total Sum of Squares (SST)
A measure of total deviation based on the error of using the sample mean yˉ to predict the observed values yi, computed as SST=∑(yi−yˉ)2.
Sum of Squares Due to Regression (SSR)
A measure of how much the predicted values y^i deviate from the sample mean yˉ, defined as SSR=∑(y^i−yˉ)2, where SST=SSR+SSE.
Coefficient of Determination (r2 or R2)
The ratio of the sum of squares due to regression to the total sum of squares, computed as r2=SSTSSR=1−SSTSSE, indicating the proportion of variability in the dependent variable explained by the regression relationship.
Conditions for Valid Inference in Regression
The two required assumptions regarding the error term ϵ: (1) for any combination of independent variable values, the error population is normally distributed with a mean of 0 and constant variance; and (2) the error values are statistically independent.
Residual Plot Error Term Violations
The four common violations of regression error assumptions observed in residual plots: (a) non-constant spread, (b) nonlinear pattern, (c) non-normal residuals, and (d) non-independent residuals.

Multicollinearity
The condition of high correlation among independent variables in a multiple regression model, which inflates parameter standard errors and reduces estimation precision.
Dummy Variable
An artificial indicator variable coded as 0 or 1 used to represent categorical independent variables in a regression equation. A categorical factor with k levels requires k−1 dummy variables.
Quadratic Regression Model
A curvilinear regression model that includes a squared independent variable term alongside the linear term, expressed as y=β0+β1x+β2x2+ϵ.
Curvilinear Shapes in Quadratic Regression
The parametric curvature shapes of y=β0+β1x+β2x2: (a) increasing when β1>0,β2>0; (b) decreasing when β1<0,β2<0; (c) concave down with a maximum when β1>0,β2<0; and (d) concave up with a minimum when β1<0,β2>0.

Interaction
A condition where the relationship between the dependent variable and one independent variable depends on the specific value of a second independent variable, represented by a product term x1x2.
Piecewise Linear Regression Model
A modeling technique that joins two separate linear regression segments at a specific threshold value where the slope of the relationship changes.
Knot (Breakpoint)
The specific value of the independent variable, denoted x(k), at which the functional slope or relationship between x and y changes in a piecewise linear model.
Variable Selection Procedures
Systematic methods for identifying the best subset of independent variables: Stepwise Regression, Forward Selection, Backward Elimination (iterative methods), and Best-Subsets Regression (evaluating all subsets).
Overfitting
The error of creating an excessively complex regression model that fits sample idiosyncrasies, causing it to overperform on training data but underperform on population validation data.
Cross-Validation
A validation technique, such as the holdout method, that evaluates regression model performance on a separate validation dataset not used to train the model.
Confidence Interval (Regression Mean Value)
An interval estimate of the expected mean value of the dependent variable for given independent variable values, defined as y^±tα/2sy^.
Prediction Interval (Regression Individual Value)
An interval estimate of an individual realization of the dependent variable for given independent variable values, computed as y^±tα/2sy^2+n−p−1SSE.
Raw Data
Data that has not yet been processed, cleansed, or prepared for analysis; also called source data or primary data.
Data Wrangling
The process of cleaning, transforming, and managing raw data so it is accurate, reliable, accessible, and prepared for analysis; also known as data munging or data remediation.
Discovery
The first step of data wrangling, in which an analyst becomes familiar with the raw data and conceptualizes its intended use.
Structuring
The second step of data wrangling, involving the arrangement of raw data into a structured layout to facilitate analysis.
Cleaning
The third step of data wrangling, focusing on identifying and correcting errors, outliers, duplicate records, and handling missing data.
Enriching
The fourth step of data wrangling, involving subsetting, supplementing by appending fields from other files, and enhancing data via systematic transformations.
Validating
The fifth step of data wrangling, verifying that the wrangled data accurately represent the raw data and that all definitions are applied consistently.
Publishing
The final step of data wrangling, which entails storing the final dataset along with its data dictionary in an accessible location for end users.
Flat File
A common pattern of structured data arranged as a rectangular table, where each row represents an observation or record and each column represents a variable or field.
Delimiter
A character (such as a comma, tab, or space) used in raw text files to separate fields within an observation or record.
Stacked Data
A data layout where all measured values for a given variable are stored in a single field or column.
Unstacked Data
A data arrangement where values corresponding to different categories of one variable are placed in separate fields or columns.
Legitimately Missing Data
Data entries that are absent for an appropriate and expected contextual reason, requiring no remedial correction.
Missing Completely at Random (MCAR)
A missing data category where the probability of a value being missing is completely independent of the missing value itself and of any other variable in the dataset.
Missing at Random (MAR)
A missing data category where the likelihood of a value being missing is related to the value of another observed variable in the dataset, but not to the missing value itself.
Missing Not at Random (MNAR)
A missing data category where the probability of a value being missing is directly related to the value of the missing data itself, introducing bias if ignored.
Imputation
The systematic replacement of missing values with reasonable estimates, such as substituting the mean or median for MCAR values or using a predictive model for MAR values.
Data Dictionary
A comprehensive reference document accompanying a dataset that describes field names, definitions, units of measure, raw sources, and cross-table relationships.
Lower-Tail Hypothesis Test
A one-tailed test where the alternative hypothesis posits that the population parameter is strictly less than the hypothesized value (H0:μ≥μ0 vs. Ha:μ<μ0).
Two-Tailed Hypothesis Test
A directional-neutral statistical test that determines whether a population parameter differs in either direction from a hypothesized value (H0:μ=μ0 vs. Ha:μ=μ0).
Level of Significance (α)
The maximum acceptable risk or probability of committing a Type I error (rejecting a true null hypothesis).
p-Value
The probability of obtaining a test statistic at least as extreme as the observed sample value assuming the null hypothesis H0 is true; H0 is rejected if p-value≤α.