Mixing matrix glitches

[mbojan]

Following recent updates to network::mixingmatrix.network() the egodata method needs:

• proper handling of NAs, e.g.:

d> data(emon)
d> mixingmatrix(emon$Texas, "Formalization")
      To
From   0.5   1   2   3   4 <NA> Sum
  0.5    0   6   3   1   1    1  12
  1      4  36  17  20   5    8  90
  2      2  14   3   5   1    5  30
  3      3  18   6   3   1    0  31
  4      1   7   2   2   0    1  13
  <NA>   1   3   4   2   0    0  10
  Sum   11  84  35  33   8   15 186
d> mixingmatrix(as.egodata(emon$Texas), "Formalization")
    0.5  1  2  3  4
0.5   0 10  5  4  2
1    10 72 31 38 12
2     5 31  6 11  3
3     4 38 11  6  3
4     2 12  3  3  0
Note:  Marginal totals can be misleading for undirected mixing matrices.

• throw informative error if attrname is not in the data

d> mixingmatrix(as.egodata(emon$Texas), "no-such-attribute")
 Error in `[[<-.data.frame`(`*tmp*`, attrname, value = integer(0)) : 
  replacement has 0 rows, data has 25 

[krivit]

emon is directed, so this isn't the best example. However, the other issues are easily fixed.