Clem: Actually the estimated value of NAT = const + b1 if there is teacher training and no other treatment, and so forth. Perfect multicollinearity would occur if (1) all values of any one variable equal either 0 or 1 (maybe student-to-textbook ratio) or (2) if some of the other treatments add to one (for example, if every school has either teacher training or principal training but not both). In either case, you must do what GRETL does—leave out a variable. Wilson Mixon -----Original Message----- From: gretl-users-bounces(a)ricardo.ecn.wfu.edu on behalf of Clement L. Rasul Sent: Sat 2/18/2006 12:50 AM To: gretl-users(a)ricardo.ecn.wfu.edu Subject: [Gretl-users] Multicollinearity Can anyone help me. I am a new user of gretl. I am doing OLS regression using multiple dummy independent variables. How do I deal with multicollinearity? Gretl refuses to calculate because of perfect collinearity. Is my approach wrong in doing multiple dummy independent variables? The equation specification that i am doing is: NAT = const + b1 Dum1 + b2 Dum2 + b3 Dum3 + b4 Dum4 where: NAT is the result of achievement tests; const is the intercept, the value of NAT if there are no project interventions; b1 is the value of NAT if there is teacher traning (dummy=1); b2 is the value of NAT if there is principal training (dummy=1); b3 is the value of NAT if there is 1:1 student to textbook ratio (dummy=1); and b4 is the value of NAT if the classroom is conducive to learning (dummy =1 if it conforms to stated classroom standards) Thanks! Clem _______________________________________________ Gretl-users mailing list Gretl-users(a)ricardo.ecn.wfu.edu http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users