for detecting multicollinearity, i know that F test is significant but t tests won’t be. my question is this - if the coefficient’s p value is .98 and the two independent value’s p values are < .01, and no significance requirement is given, we can still conclude MC even thought the coefficient’s t-test / p value is significant?
If your p values are < 0.01, it means given any significant level less than 1%, you cannot reject the null hypothesis. So, if you are asked to evaluate at 98% confident level, you can still reject the null hypothesis.
thanks eltia, thats exactly what i already know about p values. the thing is that no confidence / significance level was indicated. also, by the coefficient term i meant the constant term i.e. if you have an ANOVA table with 2 independent variables, then by the coefficient term i mean the intercept term. that term had a p value of .98 so i wasn’t sure if a significance level was not indicated in the question, could we still consider the t tests for the variables insignificant considering the p value of the intercept term, knowing that the p values of the independent variables are not significant.
I believe even if your b0 is significant (it passes the t-test), your other two independent variables (b1 and b2) do not pass t-test, but the F-test says significant. It is still a good sign that there are multicollinearity. Recall that the null hypothesis of F-test only tests the independent variables (but not the intercept). So I believe you should still reject the model based on multicollinearity.
okie dokie