R help - Nov 2008 - question about contrast in R for multi-factor linear regression models?

Hi all,

I am using "lm" to fit some anova factor models with interactions.

The default setting for my unordered factors is "treatment". I
understand the resultant "lm" coefficients for one factors, but when
it comes to the interaction term, I got confused.
> options()$contrasts
        unordered           ordered
"contr.treatment"      "contr.poly"

Here is my question:

Factor A has 6 levels, B has 2 levels,
> levels(dd$A)=c("A1", "A2", "A3",
"A4", "A5", "A6")
> levels(dd$B)=c("b1", "b2")

My question is how to interpret the resultant coefficients. What are
the bases of "dd$AA2:dd$Bb2" and "dd$AA3:dd$Bb2", etc. ?

I am having a hard time to understand the result and making sense out
of the numbers...

Please help me ! Thank you!
> zz=lm(formula = (dd$Y) ~ dd$A * dd$B)
> summary(zz)
Call:
lm(formula = dd$Y~ dd$A * dd$B)

Residuals:
     Min       1Q   Median       3Q      Max
-1.68582 -0.42469 -0.02536  0.20012  3.50798

Coefficients:
              Estimate Std. Error t value Pr(>|t|)
(Intercept)    4.40842    0.40295  10.940 5.34e-13 ***
dd$AA2         0.11575    0.56986   0.203   0.8402
dd$AA3         0.01312    0.56986   0.023   0.9818
dd$AA4        -0.06675    0.56986  -0.117   0.9074
dd$AA5         0.10635    0.56986   0.187   0.8530
dd$AA6         0.11507    0.56986   0.202   0.8411
dd$Bb2        -0.58881    0.56986  -1.033   0.3084
dd$AA2:dd$Bb2  0.26465    0.80590   0.328   0.7445
dd$AA3:dd$Bb2  0.40984    0.80590   0.509   0.6142
dd$AA4:dd$Bb2 -0.02918    0.80590  -0.036   0.9713
dd$AA5:dd$Bb2  0.35574    0.80590   0.441   0.6616
dd$AA6:dd$Bb2  1.55424    0.80590   1.929   0.0617 .
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05
'.' 0.1 ' ' 1

Residual standard error: 0.8059 on 36 degrees of freedom
Multiple R-squared: 0.2642,     Adjusted R-squared: 0.03934
F-statistic: 1.175 on 11 and 36 DF,  p-value: 0.3378

Hi losemind,
>> I understand the resultant "lm" coefficients for one factors,
but when it
>> comes to the 
>> interaction term, I got confused.
Yes, it is possible to lose your mind on this (so perhaps get a real name).
A good friend here is

?dummy.coef

In your case (i.e. treatment contrasts), your reference level for the
interaction terms are the reference levels of the factors themselves. In
your example, these seem to be A1 and b1. Assuming they are, the coefficient
for, say, dd$AA3:dd$Bb2 is worked out relative to them.

It also helps to have more descriptive names for your factors and factor
levels. This is why I haven't worked out what yours might be. In a busy day,
yours seem to be a nightmare.

HTH, Mark.


losemind wrote:
> 
> Hi all,
> 
> I am using "lm" to fit some anova factor models with
interactions.
> 
> The default setting for my unordered factors is "treatment". I
> understand the resultant "lm" coefficients for one factors, but
when
> it comes to the interaction term, I got confused.
> 
>> options()$contrasts
>         unordered           ordered
> "contr.treatment"      "contr.poly"
> 
> Here is my question:
> 
> Factor A has 6 levels, B has 2 levels,
> 
>> levels(dd$A)=c("A1", "A2", "A3",
"A4", "A5", "A6")
>> levels(dd$B)=c("b1", "b2")
> 
> 
> My question is how to interpret the resultant coefficients. What are
> the bases of "dd$AA2:dd$Bb2" and "dd$AA3:dd$Bb2", etc.
?
> 
> I am having a hard time to understand the result and making sense out
> of the numbers...
> 
> Please help me ! Thank you!
> 
>> zz=lm(formula = (dd$Y) ~ dd$A * dd$B)
>> summary(zz)
> 
> Call:
> lm(formula = dd$Y~ dd$A * dd$B)
> 
> Residuals:
>      Min       1Q   Median       3Q      Max
> -1.68582 -0.42469 -0.02536  0.20012  3.50798
> 
> Coefficients:
>               Estimate Std. Error t value Pr(>|t|)
> (Intercept)    4.40842    0.40295  10.940 5.34e-13 ***
> dd$AA2         0.11575    0.56986   0.203   0.8402
> dd$AA3         0.01312    0.56986   0.023   0.9818
> dd$AA4        -0.06675    0.56986  -0.117   0.9074
> dd$AA5         0.10635    0.56986   0.187   0.8530
> dd$AA6         0.11507    0.56986   0.202   0.8411
> dd$Bb2        -0.58881    0.56986  -1.033   0.3084
> c  0.26465    0.80590   0.328   0.7445
> dd$AA3:dd$Bb2  0.40984    0.80590   0.509   0.6142
> dd$AA4:dd$Bb2 -0.02918    0.80590  -0.036   0.9713
> dd$AA5:dd$Bb2  0.35574    0.80590   0.441   0.6616
> dd$AA6:dd$Bb2  1.55424    0.80590   1.929   0.0617 .
> ---
> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05
'.' 0.1 ' ' 1
> 
> Residual standard error: 0.8059 on 36 degrees of freedom
> Multiple R-squared: 0.2642,     Adjusted R-squared: 0.03934
> F-statistic: 1.175 on 11 and 36 DF,  p-value: 0.3378
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 
> 
-- 
View this message in context:
http://www.nabble.com/question-about-contrast-in-R-for-multi-factor-linear-regression-models--tp20413510p20416877.html
Sent from the R help mailing list archive at Nabble.com.