Considering a regression problem A x = b where the observation matrix A is underdetermined; it was noticed as regards a standard least-squares regression, in which ∥ r ∥ 2 is minimized where r = A x − b is the residual vector.
Consider a regression with outcome Y and covariates X and X.
Let us consider a regression problem with a response variable Y and two predictor variables X1 and X2.
Consider a regression model with exposure X and true confounder Z, that is, Instead of the true confounder Z, we observe an error-prone proxy variable V.
Considering a regression problem where P 1, P 2,…, P k are predictors and B 1, B 2,…, B c are responses, a multi-response regression (Segal, 1992 ) is constructed using N learning cases < p (i ), b (i ) > (i = 1, 2,…, N ).
We considered a quadratic regression significant only when its quadratic term (x2) was also significant; otherwise we considered the linear regression to be a better model.
Heatwole and Rose (2013) proposed a specific reduced-form model based on U.S. earthquakes that considered a linear regression model of predictor variables to predict losses.
We considered a specific regression model (whether simple or multiple) as the most adequate one if its AIC was at least 7 points lower than that of the other models run for a given community indicator and landscape gradient.
We considered a logistic regression to account for covariates characterizing the sequences, such that: with (Kuan and Chiang 2012).
We considered a cubic regression to be a reasonable stopping point, after which the general shape of the relations of variances with weeks did not change qualitatively.
