R help - Jan 2010 - Simple effects with Design / rms ols() function

Hi everyone,
I'm having some difficulty getting "simple effects" for the ols()
function in the rms package. The example below illustrates my
difficulty -- I'll be grateful for any help.

#make up some data
exD <- structure(list(Gender = structure(c(1L, 2L, 1L, 2L, 1L, 1L, 1L,
2L, 1L, 2L, 2L, 2L, 1L, 2L), .Label = c("F", "M"), class =
"factor"),
    UCS = c(3.11111111111111, 3.5, 3.77777777777778, 2.875, 3.55555555555556,
    2.44444444444444, 2.57142857142857, 3.11111111111111, 3.28571428571429,
    3.11111111111111, 2.77777777777778, 2.77777777777778, 3.22222222222222,
    2.77777777777778), GPA = c(3.4, 3.1, 2.9, 2.6, 2.7, 3.2,
    3.1, 3.7, 2.8, 3, 2.2, 3.3, 3.4, 3.9)), .Names = c("Gender",
"UCS", "GPA"), class = "data.frame", row.names =
c(NA, -14L))

## here is how I usually get simple effects with lm()

contrasts(exD$Gender) # check which is the reference group (reference group = F)

m.lm.f <- lm(GPA ~ Gender*UCS, data=exD) # run the model. UCS
coefficient is for females
summary(m.lm.f)

contrasts(exD$Gender) <- contr.treatment(2, base=2) # set reference
group to male

m.lm.m <- lm(GPA ~ Gender*UCS, data=exD) # run the model. UCS
coefficient is for males
summary(m.lm.m)

## try to do the eqivelent with ols ##

library(rms)
dd <- datadist(exD)
options(datadist="dd")

dd$limits["Adjust to", "Gender"] <- "F"
(m.ols.f <- ols(GPA ~ Gender*UCS, data=exD)) # run the model with
gender adjusted to f

dd$limits["Adjust to", "Gender"] <- "M"
(m.ols.m <- ols(GPA ~ Gender*UCS, data=exD)) # run the model with
gender adjusted to m. UCS coefficient is the same.

# Summary gives me results consistent with the lm results, but I want
the actual coefficients
summary(m.ols.f)
summary(m.ols.m)

#OK, so I could do
(m.ols.f1 <- ols(GPA ~ UCS, data=exD, subset=Gender=="F")) #but
this
uses df=5 instead of df=10.

Question: How can I find simple effects of UCS for males and for females?

Thanks,
Ista


-- 
Ista Zahn
Graduate student
University of Rochester
Department of Clinical and Social Psychology
http://yourpsyche.org

Ista Zahn wrote:
> Hi everyone,
> I'm having some difficulty getting "simple effects" for the
ols()
> function in the rms package. The example below illustrates my
> difficulty -- I'll be grateful for any help.
> 
> #make up some data
> exD <- structure(list(Gender = structure(c(1L, 2L, 1L, 2L, 1L, 1L, 1L,
> 2L, 1L, 2L, 2L, 2L, 1L, 2L), .Label = c("F", "M"),
class = "factor"),
>     UCS = c(3.11111111111111, 3.5, 3.77777777777778, 2.875,
3.55555555555556,
>     2.44444444444444, 2.57142857142857, 3.11111111111111, 3.28571428571429,
>     3.11111111111111, 2.77777777777778, 2.77777777777778, 3.22222222222222,
>     2.77777777777778), GPA = c(3.4, 3.1, 2.9, 2.6, 2.7, 3.2,
>     3.1, 3.7, 2.8, 3, 2.2, 3.3, 3.4, 3.9)), .Names = c("Gender",
> "UCS", "GPA"), class = "data.frame",
row.names = c(NA, -14L))
> 
> ## here is how I usually get simple effects with lm()
> 
> contrasts(exD$Gender) # check which is the reference group (reference group
= F)
> 
> m.lm.f <- lm(GPA ~ Gender*UCS, data=exD) # run the model. UCS
> coefficient is for females
> summary(m.lm.f)
> 
> contrasts(exD$Gender) <- contr.treatment(2, base=2) # set reference
> group to male
> 
> m.lm.m <- lm(GPA ~ Gender*UCS, data=exD) # run the model. UCS
> coefficient is for males
> summary(m.lm.m)
> 
> ## try to do the eqivelent with ols ##
> 
> library(rms)
> dd <- datadist(exD)
> options(datadist="dd")
> 
> dd$limits["Adjust to", "Gender"] <- "F"
> (m.ols.f <- ols(GPA ~ Gender*UCS, data=exD)) # run the model with
> gender adjusted to f
> 
> dd$limits["Adjust to", "Gender"] <- "M"
> (m.ols.m <- ols(GPA ~ Gender*UCS, data=exD)) # run the model with
> gender adjusted to m. UCS coefficient is the same.
> 
> # Summary gives me results consistent with the lm results, but I want
> the actual coefficients
> summary(m.ols.f)
> summary(m.ols.m)
> 
> #OK, so I could do
> (m.ols.f1 <- ols(GPA ~ UCS, data=exD, subset=Gender=="F"))
#but this
> uses df=5 instead of df=10.
> 
> Question: How can I find simple effects of UCS for males and for females?
> 
> Thanks,
> Ista
> 
> 
Don't use contrasts and don't change reference cells.  Just use 
summary.rms or contrast.rms with ols.  The latter gives you the most 
flexibility and is completely independent of coding of variables.

Frank

-- 
Frank E Harrell Jr   Professor and Chairman        School of Medicine
                      Department of Biostatistics   Vanderbilt University