Linked from

The 30 pages that link to Linear regression, each with the reason it gives.

• Least squaresBroader topic: Ordinary linear regression estimates coefficients by minimizing squared residuals.

• Normal distributionRelated: Normal error assumptions support familiar inference for regression coefficients.

• Linear independenceRelated: Dependent predictor columns make coefficients non-unique, a problem called multicollinearity.

• SlopeRelated: Its fitted line’s slope estimates how one variable changes with another.

• Gradient descentRelated: Its squared-error parameters can be fitted by minimizing an objective with gradient descent.

• Linear combinationRelated: Its predictions combine predictor contributions using fitted coefficients.

• Statistical modelBroader topic: It models continuous outcomes with a simple, interpretable relationship between predictors and response.

• Orthogonal projectionBroader topic: Ordinary least-squares regression projects observed responses onto the predictors' span.

• StandardizationRelated: Standardized coefficients express predictor effects in standard-deviation units.

• Calibration curveRelated: It estimates the line used to convert response into concentration.

• System of linear equationsRelated: Least-squares estimates satisfy the normal equations, a linear system.

• Neural networkCompared with: It is a simpler model whose limits highlight the value of learned nonlinear representations.

• Invertible matrixRelated: The normal-equation solution uses an inverse when the design matrix has full column rank.

• Parameter estimationBroader topic: Its coefficients are commonly estimated from paired predictor and outcome data.

• Logistic regressionCompared with: Unlike logistic regression, it can predict values outside the probability range.

• Time series analysisCompared with: Ordinary regression assumptions can fail when successive observations are temporally dependent.

• Degrees of freedomRelated: Residual degrees of freedom depend on observations and independent fitted coefficients.

• Gaussian distributionRelated: Gaussian error assumptions yield familiar estimation and uncertainty formulas.

• Ordinary least squaresNarrower topic: OLS is the standard coefficient-fitting method for linear regression.

• Scatter plotRelated: A fitted line summarizes a roughly linear pattern among plotted points.

• Student's t-distributionRelated: Under standard assumptions, regression coefficient estimates are tested using t-distributed statistics.

• Linear functionRelated: It estimates a linear trend when real measurements do not lie exactly on a line.

• Ridge regressionNarrower topic: Ridge modifies this model's least-squares fitting objective.

• Poisson regressionCompared with: Unlike ordinary linear regression, Poisson regression targets counts and constrains fitted means to be positive.

• Arbitrage pricing theoryRelated: Empirical APT applications estimate factor exposures using regressions of returns.

• Gauss–Markov theoremRelated: The theorem emerged as a foundational result about regression coefficient estimation.

• Fundamental theorem of linear algebraBroader topic: Its least-squares fit projects data onto a matrix’s column space.

• Parametric statisticsBroader topic: With specified error assumptions, it is a central parametric model for relationships between variables.

• Y-interceptRelated: The fitted line’s y-intercept estimates the response when the predictor is zero.

• Cochran's theoremRelated: Regression residual and fitted sums of squares follow Cochran-style decompositions under Gaussian errors.