R help - Mar 2007 - how to get "lsmeans"?

Dear all, 
      
    I search the mail list about this topic and learn that no simple way is
available to get "lsmeans" in R as in SAS.
    Dr.John Fox and Dr.Frank E Harrell have given very useful information about
"lsmeans" topic.
    Dr. Frank E Harrell suggests not to think about lsmeans, just to think about
what predicted values wanted
    and to use the predict function. However, after reading the R help file for
a whole day, I am still unclear how to do it.
    Could some one give me a hand? 
 
for example:
  
A,B and C are binomial variables(factors); d is a continuous variable ;
The response variable Y is  a continuous variable too.  

To get lsmeans of Y according to A,B and C, respectively, in SAS, I tried  
proc glm data=a;  
 class A B C;  
 model Y=A B C d;  
 lsmeans A B C/cl;  
run;  

In R, I tried this:  
 library(Design)  
 ddist<-datadist(a)  
 options(datadist="ddist")  
 f<-ols(Y~A+B+C+D,data=a,x=TRUE,y=TRUE,se.fit=TRUE)  

then how to get the "lsmeans" for A, B, and C, respectively with
predict function?

 

Best wishes 
yours, sincerely 
Xingwang Ye    
PhD candidate     
Research Group of Nutrition Related Cancers and Other Chronic Diseases      
Institute for Nutritional Sciences,  
Shanghai Institutes of Biological Sciences,     
Chinese Academy of Sciences     
P.O.Box 32     
294 Taiyuan Road     
Shanghai 200031     
P.R.CHINA

I verified the result from the following with output from JMP 6 on the
same data (don't have SAS: don't need it):

set.seed(631)
n <- 100
dat <- data.frame(y=rnorm(n), A=factor(sample(1:2, n, replace=TRUE)),
                  B=factor(sample(1:2, n, replace=TRUE)),
                  C=factor(sample(1:2, n, replace=TRUE)),
                  d=rnorm(n))
fm <- lm(y ~ A + B + C + d, dat)
## Form a data frame of points to predict: all combinations of the
## three factors and the mean of the covariate.
p <- data.frame(expand.grid(A=1:2, B=1:2, C=1:2))
p[] <- lapply(p, factor)
p <- cbind(p, d=mean(dat$d))
p <- cbind(yhat=predict(fm, p), p)
## lsmeans for the three factors:
with(p, tapply(yhat, A, mean))
with(p, tapply(yhat, B, mean))
with(p, tapply(yhat, C, mean))

Andy 

From: Xingwang Ye
> 
> Dear all, 
>       
>     I search the mail list about this topic and learn that no 
> simple way is available to get "lsmeans" in R as in SAS.
>     Dr.John Fox and Dr.Frank E Harrell have given very useful 
> information about "lsmeans" topic.    
>     Dr. Frank E Harrell suggests not to think about lsmeans, 
> just to think about what predicted values wanted
>     and to use the predict function. However, after reading 
> the R help file for a whole day, I am still unclear how to do it.
>     Could some one give me a hand? 
>  
> for example:
>   
> A,B and C are binomial variables(factors); d is a continuous 
> variable ; The response variable Y is  a continuous variable too.  
> 
> To get lsmeans of Y according to A,B and C, respectively, in 
> SAS, I tried proc glm data=a;  class A B C;  model Y=A B C d; 
>  lsmeans A B C/cl; run;  
> 
> In R, I tried this:  
>  library(Design)
>  ddist<-datadist(a)
>  options(datadist="ddist")
>  f<-ols(Y~A+B+C+D,data=a,x=TRUE,y=TRUE,se.fit=TRUE)  
> 
> then how to get the "lsmeans" for A, B, and C, respectively 
> with predict function?
> 
>  
> 
> Best wishes
> yours, sincerely 
> Xingwang Ye    
> PhD candidate     
> Research Group of Nutrition Related Cancers and Other Chronic 
> Diseases      
> Institute for Nutritional Sciences,  
> Shanghai Institutes of Biological Sciences,     
> Chinese Academy of Sciences     
> P.O.Box 32     
> 294 Taiyuan Road     
> Shanghai 200031     
> P.R.CHINA
> 
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}