In multiple regression, each coefficient represents the effect of one predictor on the outcome, holding all other predictors constant. It’s the amount the predicted dependent variable changes for a one-unit increase in that predictor, with every other predictor kept fixed. This is a partial or conditional effect: it isolates the influence of that specific variable from the others in the model. Important complements: dummy-variable terms are used to encode categories, and their coefficients reflect differences from a reference group. The intercept is a separate term that estimates the expected outcome when all predictors are zero. Non-linear terms (like squares or interactions) introduce curvature or interactions, with their own coefficients describing those non-linear effects. The coefficients in a linear multiple regression specifically quantify the linear contribution of each predictor to the outcome, given the presence of the others.

Adversarial search models competitive environments where two players have opposing goals, such as in chess or checkers. It assumes the opponent will act to minimize your chances, and it uses the minimax principle to plan moves accordingly. In practice, you lay out a game tree of possible moves and responses. On your turn you pick the move that leads to the best outcome assuming the opponent will respond with the move that minimizes your eventual score. This step-by-step reasoning yields the minimax value for each position and guides the best move to choose. Alpha-beta pruning is a common enhancement that trims unnecessary branches to speed up the search without changing the result. The other terms aren’t strategies for planning under an opponent’s influence: combinational explosion describes the rapid growth of possibilities as a problem scales, recursion is a general technique for breaking problems into simpler ones, and brittleness refers to a system’s lack of robustness.

Data sensitivity describes how regression estimates react to imperfect data. Linear regression relies on clean data, and outliers, missing values, or measurement errors can pull the estimated line away from the true relationship, changing both the slope and intercept. So this term captures how fragile the model’s results are to data quality. The slope and intercept are simply the estimated parameters of the line, not descriptors of how sensitive those estimates are to data issues. Dummy variables, meanwhile, are just a way to encode categorical predictors and don’t address sensitivity to data quality. If you’re concerned about data sensitivity, you’d look at robustness techniques and diagnostic measures that show how influential problematic observations are.

Straight-line distance in ordinary Euclidean space is measured by the Euclidean distance. In two dimensions it follows the Pythagorean formula sqrt((x1 - x2)^2 + (y1 - y2)^2); in higher dimensions you take the square root of the sum of the squared differences across all coordinates. This represents the direct length of the line connecting the two points, regardless of any grid or axis-aligned paths. Other metrics describe distance in different ways: Manhattan distance sums the absolute differences along each axis, Chebyshev distance takes the largest single coordinate difference, and Hamming distance counts how many positions differ between equal-length vectors. So, the distance that embodies the straight-line concept is the Euclidean distance.