Question

I have a two factor vectors v1 and v2, which appear to be closely related (the entropy of each is very close to their joint entropy). Indeed, when I do table(v1,v2), I see something like this:

     v2
 v1  a2   b2   c2
 a1  0    100   0
 b1  0    0     0
 c1  0    0     0
     v2 
 v1  d2   e2   f2
 a1  0    0     0
 b1  0    0     0
 c1  0    0     0

and so on – each factor has dozens of levels, so I get plenty of lines with all 0.

How to I print a table omitting lines which have only zeros in them?

Answer 1

Using your example:

v1 <- factor(rep("a1", 100), levels = paste0(letters[1:3], 1))
v2 <- factor(rep("b2", 100), levels = paste0(letters[1:6], 2))

R> table(v1, v2)
    v2
v1    a2  b2  c2  d2  e2  f2
  a1   0 100   0   0   0   0
  b1   0   0   0   0   0   0
  c1   0   0   0   0   0   0

Then the rowSums() function will compute the row sums for use. This works because a table is a either a vector or a matrix in disguise. Note in the sequence below showing intermediate steps how we convert the row sums into a logical vector by asking if they exceed 0.

R> rowSums(tab)
 a1  b1  c1 
100   0   0 
R> rowSums(tab) > 0
   a1    b1    c1 
 TRUE FALSE FALSE 
R> tab[rowSums(tab) > 0, ]
 a2  b2  c2  d2  e2  f2 
  0 100   0   0   0   0 

The above drops the empty dimension. If you want to keep the table format, add drop = FALSE to the call, though note the extra , in there as we want all columns hence the empty argument between , ,:

R> tab[rowSums(tab) > 0, , drop = FALSE]
    v2
v1   a2  b2 c2 d2 e2 f2
  a1  0 100  0  0  0  0