Re: product of numeric variables in lm

["Harald Hoogstrate" <Hoogstrate@POINTLOGIC.NL>]

OK, three answers within 5 minutes, must be a speedrecord.

 

This one from Beth Atkinson best explains the S+ behaviour:

 

"a*b is a + b + a:b

The definition you used of D$a*D$b only includes the interaction term and not the main effects.

If you only want the interaction, use lm(y~c+a:b,data="" class=987350115-20012003>"

 

Thanks all,

Harald

 

 

 

 

-----Original Message-----
From: Harald Hoogstrate
Sent: Monday, January 20, 2003 15:53
To: s-news@lists.biostat.wustl.edu
Subject: product of numeric variables in lm

Dear all,

 

I am doing a regression with lots of numerical variables on S+ 6.1.

Now I want to include the product of two variables a and b in my regression:

 

lm(y~c+a*b,data="">

 

and I find that the result is very different from what I get if I define D$ab<-D$a*D$b and try

 

lm(y~c+ab,data="">

 

So my questions: what does * do and, more importantly, how can I include the product of two variables without first defining them in my dataframe.

 

Thanks for any help,

 

Harald