Is it multiple regression?

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Is it multiple regression?
Message posted by SBM (via 216.222.238.246) on April 28, 2001 at 4:16 PM (ET)

I want to predict which of three subtest scores is the best predictor of the composite score. Should I use a standard multiple regression? Other considerations?

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Is it multiple regression?
Message posted by Jack Tomsky (via 208.249.113.130) on May 3, 2001 at 4:33 PM (ET)

As I understand your problem, you have three subtests: X1, X2, X2, and X3. The composite test score is Z = X1 + X2 + X3. You are trying to determine which of the three subtests has the highest correlation with the composite score Z.

First, calculate the variances and covariances among the Xj. (j = 1,2,3). Denote these by Cov(Xi, Xj). When i=j, it give you the variance of Xi, Var(Xi)

Then for each Xj, calculate the sum (over i)of Cov(Xi,Xj), divided by sqrt[Var(Xj)]. The Xj which results in the largest ratio is the subtest which is most correlated with the composite sum.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.