= 0 against the hypothesis H1: ≠ 0 (the variable significantly affects Y).To test statistical significance of each factor of a regression model, it is necessary to check the hypothesis that the regression coefficient equals to zero (the variable that corresponds to this regression coefficients does not significantly affect Y): H0: = 0 against the hypothesis H1: ≠ 0 (the variable significantly affects Y).
To estimate the statistical significance of the coefficients (the importance of model factors) for linear regression factors, Student's t-criterion is used. t-statistics is calculated as a ratio of an estimated coefficient to its standard error:
Where:
. Regression coefficient estimate.
. Standard deviation of the estimate of regression parameter 
.
Standard error (standard deviation) is an approximate value of the estimated coefficient deviation from the true value caused by sample randomness. The greater is the value of the standard error, the less reliable is the estimated coefficient of an explanatory variable.
See also: