[R-sig-Epi] problems with NA's

[David Winsemius]

On Aug 11, 2010, at 4:53 AM, Dorothea Kirchner wrote:

> Hi,
>
> I try to calculate confidence intervals for SMRs using the pois.exact
> ()-function from the package "epitools". But I have some troubble  
> with the
> NAs. How can I remove the NAs out of my analysis? It doesn't work if  
> I set the
> argument: $CI<-pois.exact($SMR, na.rm = TRUE).

You will need to pre-process your data. The help page for pois.exact  
makes no indication that it has an na.rm argument or that it will pass  
arguments on to subsequent methods. There are functions for removal of  
NA values.

?na.omit
?complete.cases

(Caveat: I am not experienced with the use of this package.) It  
appears that pois.exact expects at least two arguments, "x" (=counts)  
and "pt" (=person-time), and therefore providing SMR's (= x/pt ) may  
not give you desired results. The function will provide a rate as  
output although the author appropriately calls it a "crude rate". I am  
also puzzled by the use of leading "$"'s in variable names. That is  
not a formalism with which I am aware. You didn't provide any sample  
data or show what error messages you were getting so it's not possible  
to say what other problems may exist.

-- 

David Winsemius, MD
West Hartford, CT