#1




Pvalue of constant in regression
Hello,
When running a multiple linear regression I get very low pvalues for most of the independent variables and for the model as a whole, but the pvalue of the constant remains high. What causes this? How does it influence my results? How it can be corrected? Thank you in advance for any help. Alba 
#2




Having a high pvalue on the constant, I suppose you are wondering whether to drop the constant from the model. The answer is probably best to leave it. The interest is on the parameters associated with the Xs, and not the constant.
By fitting a regression with no intercept you can end up with negative value for coefficient of determination! 
#3




Hi,
Try to remove some of the most insignificant variables first. If you remove the constant you are basically forcing the regression through the origin. The constant ought to be there for several reasons (especially since we assume E(u) = 0 and the constant "takes care of this"). If the joint hypothesis (F type test) of all coefficient betas equal to zero cannot be rejected you should consider how your model is specified and if the assumptions behind linear regression are fulfilled. Last edited by ChrisZ; 05262009 at 10:29 PM. 
Tags 
constant, pvalue 
Thread Tools  
Display Modes  

