R help - Jun 2010 - glm output format

Hello, 
 
I am running a loop to compare some residual deviances obtained from glm, with
codes:
........
 
OUT<-NULL
for (i in 1:10){ 
myglm<-glm(mat ~X1+X2+X3,family = binomial, data =myDATA)
OUT<-c(OUT,myglm$deviance) } 
 
.......
 
In the loop, X1, X2, and X3 chage with i. If X1, X2 and X3 are not highly
correlated, OUT is a vector of 10 values.
 
The problem is:  if there there is one set of X1,X2,X3 that are highly
correlated, the program stopped, with error message:
"qr.solve(qr.R(object$qr)[p1, p1]) : singular matrix 'a' in
solve."  
 
I wonder if there is a way I can skip the highly correlated case and get a NA as
the response, instead of letting the whole loop stop with no OUT output at all. 
That is, I want to get the OUT vector with 9 values and one NA. 
 
Is there a way I can do so? 
 
Many Thanks,
Barbara
 
 
 


      
	[[alternative HTML version deleted]]

Look at help("tryCatch") and help("try").

Hope this helps a little

Allan

On 05/06/10 04:21, Bojuan Zhao wrote:
> Hello,
>   
> I am running a loop to compare some residual deviances obtained from glm,
with codes:
> ........
>   
> OUT<-NULL
> for (i in 1:10){ 
> myglm<-glm(mat ~X1+X2+X3,family = binomial, data =myDATA)
> OUT<-c(OUT,myglm$deviance) }
>   
> .......
>   
> In the loop, X1, X2, and X3 chage with i. If X1, X2 and X3 are not highly
correlated, OUT is a vector of 10 values.
>   
> The problem is:  if there there is one set of X1,X2,X3 that are highly
correlated, the program stopped, with error message:
"qr.solve(qr.R(object$qr)[p1, p1]) : singular matrix 'a' in
solve."
>   
> I wonder if there is a way I can skip the highly correlated case and get a
NA as the response, instead of letting the whole loop stop with no OUT output at
all.  That is, I want to get the OUT vector with 9 values and one NA.
>   
> Is there a way I can do so?
>   
> Many Thanks,
> Barbara