Page 1 of 1
non-linear models
Posted: Sun Apr 12, 2009 5:53 pm
by carlacoga325
hi im working on the relationship between inflation and unemployment to explain the non-linear models topic. I want to use a regression from eveiws but i dont know how to do a polinomial regression starting from the data or if it is necesary to express the ecuation in a linear form first?
Also, if i start with a lineal regression is there some test to identify whether a non linear regression would be more adequate or is the scatter analysis sufficient?
thnx!
Re: non-linear models
Posted: Sun Apr 12, 2009 6:16 pm
by startz
hi im working on the relationship between inflation and unemployment to explain the non-linear models topic. I want to use a regression from eveiws but i dont know how to do a polinomial regression starting from the data or if it is necesary to express the ecuation in a linear form first?
Also, if i start with a lineal regression is there some test to identify whether a non linear regression would be more adequate or is the scatter analysis sufficient?
thnx!
A polynomial regression is easy
You can always test if the coefficients on the higher order terms are jointly significant.
Re: non-linear models
Posted: Mon Apr 13, 2009 1:04 pm
by carlacoga325
thanx i did that and it worked, im only having trouble cuz it accepts the expression ln but not log...so it wont accept the expression y =(-0.7)ln(x) + 2.5872 and also for example when i put there the expression y =(2356.1)x^2 - (165.37)x + 7.6965 it doesnt do it becuause of things like "= is not defined".
thank u so much for your help :D
Re: non-linear models
Posted: Mon Apr 13, 2009 1:44 pm
by EViews Gareth
You might want to rephrase that question so that it makes sense.
Re: non-linear models
Posted: Mon Apr 13, 2009 3:35 pm
by startz
I can't understand what you're doing. (And if Gareth can't understand, you're really unclear.
) Perhaps you can explain more slowly, posting
exactly what you've done and what you are hoping to accomplish.
Re: non-linear models
Posted: Mon Apr 13, 2009 10:18 pm
by samijo
I guess you need to use the multiplication sigh so the expression looks like y =(-0.7)*ln(x) + 2.5872.
So in the command line you might type
ls y c ln(x)
ls y c x x^2
I don't find any problem with the ln and log function, however, you should notice that log(x) and ln(x) give the natural log. In order to get the log to the base 10 you need to use log10(x).