Hi, In the question 8 on page 212(Schweser Book1), I really cannot catch the concept of p-value of the Breusch-Pagan test. How I can determine the existence of multocollinearity, which can be deducted from p-value. Is there anybody can help me this question? I actually got stuck
With MultiC you won’t have low p values for the independents, you will think you have a solid fit though because you have a high r2, hence the problem. With a solid fit you expect at a minimum to have at least one x to have a low p value. If you observe a high adjusted r squared and no corresponding low p values for all you independents, you have multi. With multi, you have a correltaion between two or more independents, leading to a higher probability of a type two error, we fail to reject a false Ho. t-test show no independents with significantly different value than zero, yet the f-test is significant with a high r2. the f test msr/mse tells us the ratio of mean explained over mean errors. Corrective action is stepwise, in other words remove independents and see what happens. Stat software will do this, you probably only need to know that this has to be done, not actually do it. Don’t have schwesher this year. I thought you used BP for hetero? BP is n x r squared, observations times r2. with k(independents) df. You then have to look it up in a table to see. Dave