R help - Nov 2013 - if, apply, ifelse

Hello:

This seems like an obvious question, but I am having trouble answering it.
I am new to R, so I apologize if its too simple to be posting. I have
searched for solutions to no avail.

I have data that I am trying to set up for further analysis ("training
data"). What I need is 12 groups based on patterns of 4 variables. The
complication comes in when missing data is present. Let  me describe with
an example - focusing on just 3 of the 12 groups:

vec=c(1,1,1,1,1,1,NA,NA,1,1,0,0,1,NA,1,1,1,NA,0,0,1,NA,1,0,0,0,0,1,0,0,0,0,NA,NA,NA,NA,1,NA,0,NA,1,NA,1,NA)
> a=matrix(vec, ncol=4,nrow=11, byrow=T)
> edit(a)
      col1 col2 col3 col4
 [1,]    1    1    1    1
 [2,]    1    1   NA   NA
 [3,]    1    1    0    0
 [4,]    1   NA    1    1
 [5,]    1   NA    0    0
 [6,]    1   NA    1    0
 [7,]    0    0    0    1
 [8,]    0    0    0    0
 [9,]   NA   NA   NA   NA
[10,]    1   NA   0   NA
[11,]    1   NA    1   NA

Here are 11 individuals. I want the following groups (coded as three
separate binary variables):
Group1 - scored a 1 on col1 and multiple time
Group2 - scored a 1 on col1 but only once
Group3 - did not score a 1 in col1

This seems straightforward, except missingness complicates it.  Take
individual 5 - this person should be placed in Groups 1 AND 2 because we
don'tknow the score col2.  Same with individual 10, though the response
pattern differs.

I tried using if statements, but am running into the problem that if is not
vecotrized, and I can't seem to make if run with apply. I can use ifelse,
but its very clunky and inefficient to list all possible patterns:

(Note this is not complete of all patterns, its just an example of what
Ivebeen doing)
dd$TEST1=ifelse(is.na(d$C8W1raw),1,
(ifelse(d$C8W1raw==1 & is.na(d$C9W1raw) & is.na(d$C11AW1raw) & is.na
(d$C12AW1rraw),777899,
 (ifelse((d$C8W1raw==1 & d$C9W1raw==1)| (d$C8W1raw==1 & d$C11AW1raw==1)
|(d$C8W1raw==1 & d$C12AW1rraw==1),1,
(ifelse(d$C8W1raw==1 & ((is.na(d$C9W1raw) | d$C9W1raw==0) &
(is.na(d$C11AW1raw)
| d$C11AW1raw==0)& (is.na(d$C12AW1rraw)|d$C12AW1rraw==0)),777899,
 0)))))))

Any ideas on how to approach this efficiently?

Thanks,
Andrea

	[[alternative HTML version deleted]]

On 11/28/2013 04:33 AM, Andrea Lamont wrote:
> Hello:
>
> This seems like an obvious question, but I am having trouble answering it.
> I am new to R, so I apologize if its too simple to be posting. I have
> searched for solutions to no avail.
>
> I have data that I am trying to set up for further analysis ("training
> data"). What I need is 12 groups based on patterns of 4 variables. The
> complication comes in when missing data is present. Let  me describe with
> an example - focusing on just 3 of the 12 groups:
>...
> Any ideas on how to approach this efficiently?
>
Hi Andrea,
I would first convert the matrix "a" to a data frame:

a1<-as.data.frame(a)

Then I would start adding columns:

# group 1 is a 1 (logical TRUE) in col1 and at least one other 1
# here NAs are converted to zeros
a1$group1<-a1$col1 & (ifelse(is.na(a1$col2),0,a1$col2) |
  ifelse(is.na(a1$col3),0,a1$col3) |
  ifelse(is.na(a1$col4),0,a1$col4))
# group 2 is a 1 in col1 and no other 1s
# here NAs are converted to 1s
a1$group2<-a1$col1 & !(ifelse(is.na(a1$col2),1,a1$col2) |
  ifelse(is.na(a1$col3),1,a1$col3) |
  ifelse(is.na(a1$col4),1,a1$col4))
# here NAs are converted to 1s
a1$group3<-!ifelse(is.na(a1$col1),1,a1$col1)

and so on. It is clunky, but then you've got a clunky problem.

Jim