guys, problem here… with simple regression, the formula for determining the intercept point is : intercept point= average of dependent variable - [slope of independent variable X average of independent variable] so how does this formula work with multiple regression? would it be: intercept point= average of dependent variable - [slope of first independent variable X average of first independent variable] - [slope of second independent variable X average of second independent variable] - [slope of third independent variable X average of third independent variable] is this how it would rowk for multiple regression since there’s so many independent variables?
Hmm… Seems like it has to be. Did you try proving it? Normally, once you get more than 1 i.v. people start using the matrix formulation for calculating the coefficients.
My first impression is thta you are making your life more difficult that it has to be. The intercept is the value of y when all the independent variables take a value of 0 (isn’t it?)
olivier Wrote: ------------------------------------------------------- > My first impression is thta you are making your > life more difficult that it has to be. > > The intercept is the value of y when all the > independent variables take a value of 0 > (isn’t it?) yes… i understand that it is the value if all independent variables are zero… so if independent variables such as gdp and interest rates are zero… what would the intercept point be then ?
olivier Wrote: ------------------------------------------------------- > My first impression is thta you are making your > life more difficult that it has to be. > > The intercept is the value of y when all the > independent variables take a value of 0 > (isn’t it?) No he’s not. The intercept is estimated using all the data (as are all other parameters). Even if you had a data point where all the i.v.s were 0, that wouldn’t be the intercept (unless there was some physical reason or something that it had to be)
Can’t spot a LOS where you have to calculate the intercept by any other means though. I’m prepared to be corrected
I’m sure you don’t have any LOS that asks you to calculate multiple regression coefficients. There are tricks for doing it that nobody knows anymore and you just do the matrix inversion thing.
You’d need to use matrices to calculate the regression coefficients for MLR, though the formulae for calculating the slope coefficients are still pretty alright if you have 2 independent variables. Anyway, I was always under the idea that it more often tthan not is meaningless to examine the intercept because you are looking at a point where the explanatory variables do not explain anything at all?