R help - Dec 2009 - Generate missing data patterns

Dear R-users,

i try to generate missing values in a matrix X according to a given 
missingnes pattern R with the probabilities p per row.

X<-matrix(rnorm(3*100),ncol=3)

## indicator matrix for missingnes (1 observed, 0 missing)
R<-matrix(c(1,1,1,
            0,0,1,
            1,1,0,
            0,1,1),ncol=3,byrow=TRUE)

## probabilities for row 1, row 2, row 3 and row 4
p<-c(0.375,0.25,0.25,0.125)

## does not exactly what i want, because i get rows
## of missinges pattern which are not in R
X[rbinom(100,1,p[1])==1,R[1,]==1] <- NA
X[rbinom(100,1,p[2])==1,R[2,]==1] <- NA
X[rbinom(100,1,p[3])==1,R[3,]==1] <- NA
X[rbinom(100,1,p[4])==1,R[4,]==1] <- NA

So it would be great if i can get any advice how to do this. I also 
tried rmultinom or sample but without any success so far.

Another question is what to to if want the missing pattern R and 
simulate a certain amount of missingnes mybe about 10 %? I guess i have 
to mix the probabilities for each row in such a way to get approximatly 
the wanted missingnes in percent.

best regards

Andreas

Hi Andreas,
> ## does not exactly what i want, because i get rows
> ## of missinges pattern which are not in R
> X[rbinom(100,1,p[1])==1,R[1,]==1] <- NA
> X[rbinom(100,1,p[2])==1,R[2,]==1] <- NA
> X[rbinom(100,1,p[3])==1,R[3,]==1] <- NA
> X[rbinom(100,1,p[4])==1,R[4,]==1] <- NA
If you run these lines directly after each other,  the same rows can
be selected multiple times and result in combined missingness
patterns. At least the propabilities for missingness you defined will
be different, because for instance 111 can also be created by a
combination of two other patterns.
> Another question is what to to if want the missing pattern R and simulate a
> certain amount of missingnes mybe about 10 %?
Maybe have a look at sample()?

Coen