Dear R-users, I've been reading a bunch of things on linear models but cannot quite find a clear answer. How can one determine whether a linear model is significant or not? For background info, I am modelling the response of topographic slope to the distance of a catchment's outlet. Some guys have shown that if there is a significant fit to a linear model, one can deduce the dynamic state of the basin, that is, whether erosion is as strong as rock uplift, erosion is smaller than rock uplift, or erosion is greater than rock uplift. I am thus to test 4 situations: Situation 1: a linear model is inappropriate for describing the data, the scatter is too large, and thus a linear model is unfit to explain the data. Situation 2: the linear model of the kind "y = b0 + b1 * x" is fit to describe the data, ie data points lie close to a straight line. Situation 2a: the relationship between slope and distance is significantly positive Situation 2b: the relationship between slope and distance is significantly null (ie data is clustered around a line with b1 non-significantly different from 0) Situation 2c: the relationship between slope and distance is significantly negative I am confused as to what test I should use for discriminating these situations. The glm offers an indication about the significance of regression parameters. So in the case where b1 is significantly different from 0 (p value <=0.05 for a test where H0: b1=0; H1: b1 != 0), it is straightforward. But I don't know how to discriminate between situation 1 and situation 2 (ie whether a linear model is significant). Any suggestion are welcome Cheers, Thomas *** Le contenu de cet e-mail et de ses pi??ces jointes est destin...{{dropped}}